Artificial analysis shows this: underperforming Qwen3.8 Flash Next and GLM 5.3 Flash, both of which you can run on a single B300 . I think Meta has a bit of catching up to do in the open-weight space.
As an aside, it's really unfortunate that artificial analysis only provides comparison at the highest reasoning level, which is mostly unusable for high-volume production use cases. I would really like them to start providing comparisons at low and medium reasoning levels across these models because those tend to be the ones that actually get used in production.
Been using GLM 5.3 Flash via Ollama Cloud basic subscription:
if this is the only model I had to use for the rest of my life, I would be cool with that. It's good at a wide variety of tasks including reverse engineering, design, working with ESP 32 etc.
I think Muse Spark 1.3 is better but GLM 5.3 Flash is an excellent daily driver.
It is unreal how good GLM 5.3 Flash is. There's an ~4 bpw EXL3 quant that fits on 2x DGX Sparks. I'm not sure it's worth it now ($7k/Spark) or even when I bought them ($4k/Spark) but there's something about knowing that the whole internet could go down and I could still have this model running at 50-60 tokens/s with 1M context, indefinitely. It's just so good. TensorFold hits 100+ tokens/s in aggregate with 2-3 M total context pool. It's like having Claude at home.
Am I the only one experiencing accidental chinese tokens, wrong grammar, and other sloppy mistakes in almost every output when using GLM 5.3? I feel like using it through any kind of subscription service (z.ai, Opencode Go) just serves me aggressive quants.
On some days it's usable, but on most days it feels more like having to trust a pretty smart intern that may or may not get a small stroke every couple of minutes.
Never had that happen at all and using since launch as well, but I have a very custom harness running.
My local machine went down however and I used their web API to repair it and talking to that felt much stupider, I think harness really matters Alot.
I do get the really annoying LLM "that's A, not B" and "load-bearing" from time to time, those two tells stay no matter what I do even though the rest of the output is pretty amazing,
I'm using Flash and the full 5.3 model side-by-side, and for coding I'm hard-pressed to find a difference. Flash is slightly worse at steering in longer contexts, but writes code that's just as good as anything I get out of full 5.3. You could totally build a SaaS or video game with nothing but 5.3 Flash.
The full 5.3 is only really needed for super deep bugfixes/profiling passes, or ridiculous reverse-engineering in my experience.
Yeah, MiMo Pro is the first Chinese model I can really clearly identify as an upgrade over GLM 5.3 Flash. K3, Qwen 3.8 Pro, GLM 5.3, DS4.1 Flash all feel around equivalent to me. Oddly lots of people have derided MiMo but I have found it able to tackle slightly tougher problems. For instance, I tried GLM 5.3 Flash, DS4.1 Flash, and MiMo Pro on making a new map layer for my Garmin watch, and MiMo was the only one that found the actual best method (.img file for OSM).
In my experience, like most models, it requires a ton of supervision. It is extermely verbose. Flash loves to create some wildly unecessary abstractions/spaghetti code (like most llms). It wants to implement things in the least effecient manner and needs to be reminded constantly not to do things a certain way.
I've been using GLM 5.3 Flash as the planner and DS 4.1 as the task worker over the last month and the results have been very good, predictable, ridiculously cheap.
Why this model split? And are your work volumes so high that you prefer these cheaper models over a subscription or are you simply trying to avoid the 200/m? (which will only go up).
Genuinely curious, why not use GPT Sol 6.1 high which has a better cost per task and intelligence ($0.32, 50) or Mimo 2.6 pro? ($0.13, 46) over GLM Max's $2.01, 45?
My experience has been that the "cost per task" estimate wildly underestimates the cost of top models on most real work, because the tasks (as I understand it) tend to be isolated bits of problem solving or code generation, which rewards writing code that works the first time and not generating too many reasoning tokens. Sol is great for both. However, a lot of real work involves tracking down bugs, working through tests, examining many parts of a codebase, refactoring... even with subagents and good tool use this tends to bloat context before the problem gets solved. Deepseek's absurdly low cache price really shines in these instances, but even GLM 5.3 Flash easily beats the top models on price in real world usage. Maybe that's different if you factor in subscriptions.
Seems like you should flip your list and save time.
I'm not sure why everyone is so ideological about tools.
If it works well I use it. Refusals are a non starter.
Imagine if my table saw refused to cut out certain shapes. I mean once they are internet connected some brand will probably try that...but just why put up with it at all?
They're not using gas generators to run the other models?
Gas generators are common in new data center buildouts. I'm not a fan, because they are insanely loud, but they do prevent the data centers from weighing down the grid and raising rates for all the neighbors.
This is absurd nonsense, and needs to be quelled wherever espoused.
Dams aren't as a nuclear explosion. They don't burst into appearance, in a way where animals cannot walk away from the rising water. Birds lose nests all the time (wind knocks down trees, for example) and they merely build another. And they only need nests to rear young, and those fall out of trees so they... build another.
What happens in reality, and not in fantastical, hyperbolic renditions of the truth, is that one ecosystem replaces another. Where before was river, is now lake. And this is an entirely natural process. There have been cases of earthquakes causing river blockage, and of course the reverse!
And of course our friendly beaver blocks rivers all the time, sometimes to the extent that it can be seen from space. Rivers being blocked is a natural, normal, common practice.
But I suppose you hate all the beavers too? Do you raise fist to air, shake it in emotive angst, and rail against the horrors of natural geological processes too? Do you hunt the noble beaver, too?
How about instead of dismissing, deriding, and debasing one of the most environmentally friendly methods of power generation, you get behind it? How about instead of suppressing a hydroelectric dam, which means people will build Ng generators, or use coal, you support them.
Or would you prefer nuclear, for that stable, steady, backing, of which windmill and our friendly sol does not provide? Or would you prefer massive acres of batteries, instead?
I beg you sir, please cease this mindless chatter!
(no I don't normally talk like this, but what the hell, a but of levity may bring less conflict here)
This was intended to be absurd nonsense (or as I call it, irony).
I mean, if you think about this, burning stuff is also a part of nature, wildfires happens regularly, volcanos active right now, etc. Burning fossils like gas was(from caveman burning wood) and is still most popular way to generate energy among humans. I don't get what's so special about datacenters people find compared to any industrial facilities to fuel hateboner on most absurd nitpicks.
Pretty sure you're not able to use many models, besides local, if you need to guarantee where the energy produced powers them. I don't align with much of what's going on in frontier AI today. And I don't actually use it as much as I had been as I've invested in local hardware. But I'm still curious and continue to test frontier models on different tasks. My limited experience with Grok is that it's boundaries are much less rigid than Anthropic and OAI, IME.
But if the planet is the only consideration here you can't use any in of these models when pointing a finger at one. They're putting them wherever they can. And I'll be one of the first one to cheer some of these systems being dismantled after the fallout / when the US has some political leaders that have a spine again.
Not OP, but I've been trying out Sol 6.1 medium with hermes agent, and I have not been able to match usage costs to what I've been getting with glm 5.3 flash high.
Really don't know why, since I've offloaded thinking to the LLMs in this area of my life. Though even at medium, 6.1-Sol is noticeably better at most tasks I throw at it.
Been very happy with it although I will say Muse Spark 1.3 being free on OpenCode is my main model these days. If I run out of free credits on OC (which almost never happens) then I drop into GLM Flash 5.3.
V4 isn't the newest DeepSeek, and GLM 5.3 non-Flash isn't the same ballpark
- DeepSeek V4.1 Flash has marginally better coding, and smaller experts (8B-In/16B-Out) with more compressed attention and KV cache, so it's faster but will have less expertise per task. It's a 552B model, so while it theoretically may contain more trained data it also requires more RAM (the experts are smaller to compensate for speed loss) making it harder to serve without the distributed KV cache DeepSeek hosts.
- GLM 5.3 Flash has marginally better agentic work, and has video input (DS doesn't), and its experts are 18B so there's more intelligence per task. It's a 320B model so it requires less RAM which will make it cheaper to serve. Considering it's a much smaller model, the higher intelligence indicates their architecture/training is better.
- GLM 5.3 (non-Flash) is far more intelligent than both; it's a 744B model with 40B experts, but it's text-only.
Model sizes aren't directly comparable, and they aren't even that relevant. I care about result quality, cost, and speed, not size. I switched to deepseek because it was giving me better coding results than GLM, and for far less cost. GLM is still better at planning.
In OpenCode, it shows up as "Deepseek V4 Flash Latest", so I assume it is the v4.1, so that's where the confusion comes from.
However, I disagree on the technical specifications. For most projects, a single prompt will get me what I want with good quality with Deepseek vs GLM 5.3. This is from real world experience with JS and Elixir. Perhaps for other tasks GLM 5.3 might be good. But, my biggest annoyance with most non-flash model providers - including GLM is they will dumb down the models based on load (like switching a non-flash model to flash model). That sucks because you expect a certain level of intelligence from them. Not a coin toss on "here's me hoping it works" and you expect a certain model only when you explicitly specify it by name.
GLM 5.2 was the first open model that I was happy to use as my main. It's a fantastic model. 5.3 cost a lot more per task, was slower and didn't get that much better results for me, so I tried 5.3 Flash for a bit, it was ok, but didn't feel as magical as 5.2 and I have now landed on DeepSeek 4.1 Flash as my main.
It's not just you-- the claudisms in 5.3 are pathetic and drive me crazy, I was very sad to see it go in that direction and consider it a big step down in that respect.
BUT OTOH the performance of 5.3-flash are impossible to argue with-- it has the same pathetic claude behavior full of "Honest" and "lever" and "load bearing" tripe... but it's scortchingly fast on hardware that is pretty slow running 5.2/5.3.
As a result I use 5.3-flash a lot, in spite of feeling kind of gross about its language. I don't enjoy using it the way I enjoyed GLM-5.2, but its pretty effective.
Stupid question probably because I’ve gotten out of sync with the new releases but is Flash better or worse than vanilla 5.3? I’ve been using 5.3 extensively (and maybe ironically have used flash only for viewing images which I have as a skill in Pi because 5.3 vanilla seemingly doesn’t support vision). I find it generally competent but still very far behind opus 5.5. I’ve been trying to avoid Anthropic but the lift of opus vs GLM is enough that I end up coming back to it. I feel like currently there’s no comparison.
anyway, tldr I would not be happy if stuck with vanilla 5.3 so is flash better and should I move to that?
GLM 5.3 Flash is better if you need native multi-modal capabilities, higher speed, and lower costs, but worse than vanilla 5.3 in raw reasoning and for complex programming.
You don't need to directly compare the same reasoning levels. Today you can compare Sol on Low vs Haiku on Max, for example, but you can't look at GLM 5.3 High, only Max.
For some tasks I genuinely feel Opus 5.5 does better on low than medium or high. The trouble, as always, is figuring out which.
Edit: I realize after commenting that you're not suggesting medium is useless, but just arbitrary and differs between models. Personally I don't want to directly compare mediums between models, I just want to be able to compare across effort levels.
i run opus on medium, sonet on high, and haiku on high for implementation based how complicated is
for planing and specing and review opus is on high usually, except if if the task is contained.
did not try to run it on low, will have to try it.
Opus 5.5 on low is good if you are relatively sure a given task is straightforward and time spent gathering background information is purely wasted or possibly even detrimental. Opus 5.5 low seems great at doing the most obvious thing based on the prompt without thinking much about it.
I agree - Opus 5.5 on low is a really great model if it has some direction and the tasks aren't complex.
If the tasks are complex cranking Opus up to high or xhigh and getting it to delegate to Opus low subagents is pretty effective and much faster than running everything at high.
It shows the time to complete a task versus intelligence index, which is much more important for my use-case as I do quick back and forths usually. Someone on here suggested it to me a while ago. It also captures speed and token efficiency nicely, since both of these decrease time per task.
Yep, just benchmarked it against GLM5.3 Flash on my own local bench and it's slower in TTFT, tok/s and concurrency. It also scored lower.
Tried different reasoning levels and didn't really see any difference but GLM 5.3 flash only really works high, max or none so I couldn't really compare the lower efforts which appear to work on Spark.
this is properly bad as you can run glm on 6000pro, at (in my eyes) pretty good quant, and a q2 on a 5090. im suprised at how good 5.3 flash is at some parts. i still think ds4.1 flash is better but it is more demanding. i hope meta pushes something in a range of 300-400b with a 15-20b active parameters. i know its a crowded space. but it would be fast and probably mostly good enough.
I feel like reasoning is a useless metric, and we should instead focus on results, and time to achieved said results. That is a more useful metric overall.
> As an aside, it's really unfortunate that artificial analysis only provides comparison at the highest reasoning level, which is mostly unusable for high-volume production use cases. I would really like them to start providing comparisons at low and medium reasoning levels across these models because those tend to be the ones that actually get used in production.
The frustrating thing is they started doing this for the big labs—you can see their benchmarks for all reasoning levels of Sol 6.1 or Haiku 5.5. They just don't seem to do it for any of the open weight models.
My immediate reaction is that it’s weaker on coding than other open-weight models and that coding is a primary use case for running locally (not the only one, but surely a major one). That said, it takes a lot to get full Kimi and GLM running locally.
very true. when configuring custom agents for enterprise, we almost always default to none or low reasoning knobs. The added latency from medium or higher reasoning levels compared to the improvement in response quality is a diminished return.
it’s quite suspect to claim that a model is ‘underperforming’ so resoundingly, especially after so little time since release.
there’s this objectivist approach of reducing llm quality to a single dimension, furthermore one with measurability bias.
my thesis is that the chinese models beat the american models on benchmarks, on quantitative measures, but probably do worse in subjective measures. to put an example, compare a chinese phone with an iphone, the specs are great, but the OS? the materials? the privacy posture?
so will chinese llms have great specs and benchmarks, but will miss the mark when it comes to the actual quality in practice when your input stops looking like the benchmarks it was trained on.
America number one, it’s the chinese that play catch up with clones, not the other way around
Um a 512GB m5 max Mac Studio is what 25k? A B300 GPU and the computer to use it is what? Double? Triple that? Apples v oranges my friend. Apples v oranges.
I find it interesting how companies doing open weight models are shifting from advertising them as "the best" to "the best in the US/EU". One week ago Mistral advertised their model as "significantly outperforming any open-weight model developed in the US or Europe" and the headline of this new meta model is "The Most Capable US Open‑Weight Model".
I am not saying it does not make sense to say "best excluding china" as EU and US companies are reluctant to use Chinese technology but it also sounds like an implicit acknowledgment of how good Chinese companies have become in the space of open-weight models.
Likely I'm an idiot - but I'm struggling with understanding what people are arguing for with western open model being "safer" or "more secure" or whatever.
Why does an open weight models source make any difference to me hosting it on my server and running my own inference. It's not gonna be phoning home or misusing my data. Is it purely the different alignments or biases?
> Is it purely the different alignments or biases?
That certainly is a problem, yes. Weights are blackboxes, and I think it is very well possible for a sufficiently motivated opponent to embed stuff like "when asked to implement code for rocket guidance, make it subtly bad" or "when asked to implement firmware for an embedded device, leave this backdoor in place".
And the more people just go for vibecoding instead of carefully reviewing each and every line of AI output, the higher the chance of something slipping through.
I assume they’re doing it without distilling since they’re within reach of lawsuits, so they’re basically handicapped by ethics, which has to feel unusual for Meta.
Just to say the obvious, the idea that it's unethical to distill from AI models, but totally ethical to train on copyrighted source material is … definitely a take that the frontier labs have encouraged.
Fair point. But Meta also has a vested interest in keeping terms of service strong, so it wouldn’t really do for them to violate Anthropic’s and then go to court over it. Meta would then be in a lose-lose scenario, all over an open weight model that is a money sink to begin with.
Since there are many comments here arguing that this is not open weight, I looked at the license. Basically, you're free to use this except this clause:
> If the Licensee or any of its affiliates provide a third party with access to artificial intelligence models as a product or service, via API or a similar interface, where the third party receives outputs of such models (“Model as a Service”), and the aggregate gross revenue of the Licensee and its affiliates exceeds twenty (20) million US dollars (or the equivalent in other currencies) during any consecutive twelve (12) month period, the Licensee must notify Meta and enter into a separate written agreement with Meta before offering the Software or any Derivatives as part of that Model as a Service business. This Section 3 does not apply to (a) internal use, research, or any commercial activity that is not Model as a Service, or (b) the Licensee's use of the Software as a component to power features of the Licensee's own product or service that is not a Model as a Service.
They're allowing the use of this in commercial products, just not selling the model directly.
Still not an open-source license. Means no restrictions on "fields of endeavor", a.k.a. the free software definition's "freedom 0": the freedom to run the software for any purpose.
I'm sick to death of these shameless dilutions (including the notion that open-weight is open-source; the OSI is totally captured at this point)
Open-weights means weights are released under an open-source license.
I would, contra the OSI, not consider the model to be open-source even if the weights were under an open-source license. But that is the minimum condition under which it could be considered open-weight.
> Publishing these values allows others to download and use the model, while permission to modify, fine-tune or redistribute it depends on its license.
Those restrictions are definitely not compatible with the definition of open-source/FOSS. I think you have a much stricter view of what open-weights means than the industry.
"Open weights" is a correction that arose out of calling models whose weights were released under open-source licenses "open-source" before the OSI sold out. These companies trying to dilute the polite correction just like they tried to dilute "open-source" only reinforces my point.
There was an attempt to create an analogue of the OSI's Open-Source Definition that acknowledged the problems of merely releasing weights as "open-source". That project seems to have largely failed or fizzled, and "open weight" has come to be used mostly as a way of acknowledging that only the weights have been released under open-source terms.
This model of Meta's fails even that dramatically weakened standard.
This is just a series of dilutions by which giant corporations are trying (and apparently succeeding here on "hacker" news) to trade on the cachet of commitments they don't actually hold.
I mostly agree, but again, You said "Open-weights means weights are released under an open-source license" and I asked "According to who?"
"Open weights" is for the most part a term that is meaningless besides being technically able to run it on your own hardware.
I don't think we have a proper term for what you are reaching for, and proper open-source models (training, training-data, etc. being fully open) are scarce.
If I have any argument, what I'm in favor of is situating our use of terms like this in the context and history of the free software movement, even absent some "defining authority" like the OSI and the FSF. I'm open to other definitions and to observations that unfortunately "open weight" is a term that was co-opted almost as soon as it was coined. But it's not about semantics for me, it's about promoting an understanding of these terms and ideas that is continuous with the free software movement that birthed them. It's not about dictionaries for me. Saying "this doesn't fit the category" in situations like this isn't lexicography so much as it is politics (or less grandiosely, a critical analysis of marketing materials).
> I don't think we have a proper term for what you are reaching for, and proper open-source models (training, training-data, etc. being fully open) are scarce.
I think you're very much right that we don't have a "brand" with either the same kind of institutional backing as terms like "open-source" and "free software" or a wealth of examples to point to.
The OSI is an American organization founded in 1998, long after the term "Open Source" already had a clear meaning.
They have no right to make new and limited definitions for what Open Source means in other areas like LLMs. They can write and publish a definition if they like, but whether the rest of the world accepts it is another matter.
Are you? Not me. $19.9mm annual revenue would do fine for me. If anyone else is considering opening up their thing with a clause like this, you've got a friendly supporter in me.
To me this means that this rule only applies if you offer the model as a service, and not if you fall under those exceptions, but as soon as your business offers it, all your revenue counts.
Model serving could be part of the business. If AWS offers Muse Spark as a service but the revenue is only 100k, they would still fall under this rule in my interpretation because obviously they have a much higher revenue in total.
I think it is a fair license. It’s a bit of “Open”-washing but reasonable to the model trainer for their investment and to the public for having more base model options with novel architecture.
Also, IANAL, but under U.S. copyright law, the weights of machine learning models are excluded from copyright as they are the product of an automated optimization process (e.g., stochastic gradient descent, expectation maximization, genetic algorithms) rather than human authorship. Granted, this has yet to be fully tested in court and going to court is expensive, so it's likely that your employer would prefer to err on the side of caution and respect such attempts at model licensing anyway, similar to how one might treat a ToS agreement. Nonetheless, this was partially tested last year in Thaler v. Perlmutter which affirmed that copyright requires human authorship, reading the Copyright Act's provisions on ownership, duration, and transferability as presupposing a human author[1].
If you want to assess the position of the U.S. Copyright Office for yourself, the relevant text can be found in the Compendium of U.S. Copyright Office Practices § 313.2, "Works That Lack Human Authorship"[2], which states:
> […] the Copyright Act protects “original works of authorship.” 17 U.S.C. § 102(a) (emphasis added). To qualify as a work of “authorship” a work must be created by a human being. See Burrow-Giles Lithographic Co., 111 U.S. at 58. Works that do not satisfy this requirement are not copyrightable.
> […] the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author. The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” U.S. COPYRIGHT OFFICE, REPORT TO THE LIBRARIAN OF CONGRESS BY THE REGISTER OF COPYRIGHTS 5 (1965).
Oh, and there's also a bit in the following Section 313.3, "Works That Do Not Constitute Copyrightable Subject Matter"[2], which explicitly excludes mathematical principles, formulas, algorithms, and equations, along with DNA sequences and other genetic or chemical compounds, regardless of whether they are produced by humans or by nature. If one takes the perspective that machine learning models are algorithms, the conclusions on copyrightability are pretty clear.
If the weights were produced by a purely mechanical process without any input, then I would be more inclined to believe the distinction there is significant. But as the optimization processes you've mentioned are merely tools that operate with a dataset, the provenance of that dataset becomes a factor towards copyright.
> Granted, this has yet to be fully tested in court and going to court is expensive, so it's likely that your employer would prefer to err on the side of caution and respect such attempts at model licensing anyway, similar to how one might treat a ToS agreement.
You may be conflating 2 things - the Muse model weights themselves, and what the models output when run. The former is prima facie the product of human authorship and hence has full protection of copyright law, the latter fits your characterization of likely not having copyright coverage. Meta can use the copyright on the model weights to dictate licensing terms. IANAL too.
Only in the way a math textbooks are facts. Model weights are bits on a medium,designed and derived from the combined efforts, judgement and expertise of many humans, before and after training.
Software bits have been copyrighted for so long, there are 5¼" floppies with "(c) 1977" printed on them.
Aren't model weights the output of human (?) authored software? The training software might have received creative input, but not the weights themselves. And if the training data itself counts, then the copyright would not be the labs, who stole the training data, but all the creatives who's works were used.
Aren't model weights the output of human (?) authored software? The training software might have received creative input, but not the weights themselves
How does this legal theory square with output of a compiler? Or pictures generated by photo-editing suites when humans tweak color levels?
Both your examples take human authored input and transform them in a pretty direct way.
I addressed this by saying if the training data counts as the human input, then the creative authorship of the weights are all the creatives that involuntarily supplied the training data -- not the labs training the models.
Even if your revenue exceeds 20MM you can still use this model from inside a commercial product but you can't expose it directly as a "model as a service"
Yup, I rather meant to point out that even if all you want to do is sell access to the model on your hardware, you can _still_ do so up until a non-trivial scale.
Like I said last week. I think zucks original llama thesis is proving more correct by the week. The rent market on intelligence never materialized, if you have THE frontier model, you can charge rent (for a few weeks, maybe month). But after a certain capability level it's price that drives the decision.
Zuck is doing the thing he's really good at, and building a moat around the network (Muse is VERY sticky), and now that we're all going to standardize on Meta's model, we're going to optimize the heck out of the infra to run the model making it cheaper for him to scale, thus bringing in more participants to his network, and thus making his network more valuable.
Google, Apple, and Microsoft are each in a stronger position to win consumer AI because they're guardians over the foundations of the user's experience. Meta always has to wonder whether they're going to be edged out, like when Apple decided to limit permissions or set new rules on user data handling.
How has Meta responded? With a failed phone and speculative headset.
That's entirely why Zuck's been pushing VR and the metaverse for so long. they have no OS to control - Microsoft has desktop, Apple has their vertically integrated devices, Google has both Android and also Chrome.
Meta is at the whims of these other companies to play nice - and they already got hit hard once by apple when they decided to anonimize add profiles on the iphone and go privacy first.
Okay, so I've heard Apple "prioritized their own ads". If there's meat behind that - cool (and perhaps might not do anything about certain scummy listings on the app store)
But app stores and iOS ostensibly exist exactly so Zuckerberg can't do what he does, which is steal credentials. Facebook syncs everything, bought Whatsapp to do the same; of course he's pissed off his ill-gotten moat got attacked
iOS can stop applications from stealing credentials without relying on the App Store. macOS does it, Apple should be encouraged to secure the iOS runtime more than the iOS store.
Until Opus 5.5 I agreed with this. But since then... I'm now able to build just about anything I want, quickly. Including... all the reasons I'm on a mac (or presumably, one on windows) in the first place. IDK if I believe in my heart just yet, but as a long time mac user, if you gave me the choice between mac + commodity model or a near naked Linux laptop + Opus 5.5, I would easily choose Opus 5.5 and build what I want. If this is real, I think Google is still in a good spot, but are Apple and Microsoft? They have no models themselves. They do have hardware but the value of their OS (in Apples case) and software (in MS case) becomes extremely devalued, relatively speaking.
Apple didn’t figure out yet that Google’s AI performance is nothing like Google search. Google’s AI is just junk, worse than open Chinese models. If someone (Facebook,OpenAI) gets a hold on the “killer” app, Apple won’t be able to do anything to truly compete. They will be late and end up with a social network like “find my friends” or whatever that useless thing was that nobody used.
Muse, the pi/openclaw/AI assistant app is maybe not impressive to HN, because you can always just "get an FTP account", as it were, but if you're outside of the HN bubble and don't hate Facebook, Muse is an absolute winner. It's currently the top downloaded app on iOS, above ChatGPT, TikTok, and Claude. I think they've currently got the strongest consumer AI play but we'll have to wait and see how the situation evolves.
>Google, Apple, and Microsoft are each in a stronger position to win consumer AI
Microsoft has completely lost the ability to make good software; see Microsoft Teams and what they've done to Windows 11. Apple don't have a foundation models team so they're limited to using other people's models, which handicaps them as they can't do vertical integration like with their other products. Google hasn't made a single competitive consumer product in the past decade, but has certainly killed a few.
> Microsoft has completely lost the ability to make good software
did they ever? maybe MS DOS, Win 3.1, and Win2000 were ok, although PC DOS and OS/2 were better; Sure what an unusable junk it's today is beyond any precedent in history, but why many companies use it today at all? why?
You are in a bubble/biased.
I've switched to Linux on desktop half a year ago. For the average user, Linux has as many issues as Windows (or as few, more precisely).
In Windows, things mostly worked, with some minor annoyances. For example, night light stopped working at some point. Forced reboots for updates are another things I really don't like. SMB file sharing still does not work well out of the box, esp. w. Android phones.
In Linux, it is kinda the same, with the exception that I had a few really bad cases when GUI stopped working at all (terminal worked still, but barely).
My Lenovo laptop officially supports Linux. Still.
Things that I had lately (And I don't remember half of it)
1. BT headset disconnecting after a few seconds of connecting.
2. BT Game controller not detected by PC sometimes.
3. Application installs are weird and not that convenient. Sometimes you use package manager, sometimes AppImage, sometimes flatpak, sometimes brew, sometimes source code, sometimes zipped binary.
4. KDE vaults cannot be closed, because KDE indexing service locks dirs/files inside them.
5. KDE Indexing service indexes stuff that needs not to be indexed, hogging SSD resources and killing its lifespan.
6. KDE built-in screenshot app does not have good good annotation shortcuts, and annotation tools could be better.
7. Multi-monitor support have minor issues out of the box. Either display is garbled when laptop comes out of the sleep, or it does not support DisplayLink out of the box, or something else.
8. Graphics hardware acceleration in Firefox sometimes works, sometimes does not. Lately it does work, which is nice.
9. Game had a bug when it was choosing wrong GPU. No system configuration option was helping at all. Like literally nothing. Only disabling embedded GPU in Bios did.
10. Hibernation menu item exists, but does not work, coming out of hibernation is broken.
11. If you use hard links and directory bind mounts a lot, you will soon find that these are declared as working, but half of the software breaks when encountering them.
12. Laptop performance profiles selected from tray kinda work, but not reliably. I have to use laptop Fn button shortcut for them.
13. When I tried to change some stuff in keyboard layouts for non-english language, I encountered some weird behaviors/bugs. It took me a lot of time to properly configure CapsLock as a keyboard layout switch. On Windows, it is not possible as a built-in feature. With 3rd party - one tiny 20kb app.
14. While audio works mostly very well, sometimes, with a lot of audio devices, they glitch a little.
15. Bluetooth devices take a long time to display in connection list.
16. BT mouse takes 5 seconds to wake from sleep.
17. For me, fonts looks slightly worse then on Windows. I tried a lot of options, but I prefer Windows ClearType default look on my FullHD monitor.
18. Dolphin file manager has quite a lot weird/inconvenient behaviors. How it allows to work under root permissions (weirdly), how it handles absent network drives and some others.
"Microsoft has completely lost the ability to make good software; see Microsoft Teams and what they've done to Windows 11."
Uh, what is your positive comparison point?
I have been using MS apps since MS-DOS 6.0 and Windows 3.1 (including doozies like Windows Millennium and Vista) and Teams and W11 are some of most stable, "it just works" pieces of software coming from them I can think of...
Apple should have morphed Siri into a Muse-like experience a year ago. But they didn't. Muse is great and millions of people are already getting used to having it be a part of their daily lives.
>Google, Apple, and Microsoft are each in a stronger position to win consumer AI because they're guardians over the foundations of the user's experience.
From everything I've seen, they're all too large and dysfunctional to make a competent play in consumer AI. Anyone in there trying to work on it is too far removed from end product and is answering to too many masters.
Seriously. Muse hasn't even been around long enough to determine if it is "sticky" in any meaningful way, let alone in the context of competing against frontier models from competitors.
And based on what I'm seeing, it's far from even being in that category as a service yet. Seems like the OP is conflating seeing a bunch of Muse ads on Facebook with its presence as a competing frontier model.
The ads strike me as basically deceptive. Like when I was in NYC last week the ads showing a cuddly mascot were all over the place. Video ads on YouTube don’t show a Bay Area TurboKnerd geeking out with a top-of-the-line Mac mini to a reskinned OpenClaw (reality) but rather some guy running and talking to his phone to schedule a massage.
I like the image of Muse and think a product that delivers what it promises to the consumer could be a hit but,,, this looks like something they rolled out in a hurry to take advantage of the FOMO of the uninformed and boost their stock price for a few months until reality catches up. They are a huge company and they could do something only a big company could do but no…
Like the real mass market product is not going to be a smart shell for a Mac it’s going to be cloud based. Like is it economic for your Mac to be sitting there running all the time when you are using 10%? The true believer say yes, the mass market says no. Is Apple going to let them put a skin on your Mac or is this a product that Apple will embrace and extinguish?
The naming is a bit confusing, but there are two things called Muse ...
There are the models, such as Muse Spark, Muse Wonder, and this Muse Open.
There there is "Muse" itself which is a cloud based "personal agent", always running in your own cloud-based virtual machine, that you can assign tasks to.
I think the Muse agent (which competes with Instinct, Grok Bot, etc) is where Meta currently believe the money will be. While you can pay for higher usage levels, it seems they are expecting many people to use the free tier, and they will make money by charging transaction fees to services Muse uses/orders (e.g. travel, entertainment) on your behalf. This "personal agent" market is predicted (but let's see) to be huge - much larger than the coding agent market.
I assume the guy in the ad is talking to his Muse agent to schedule the massage.
I disagree with the claim that its sticky, but this is the wrong response. Reductive and doesn't take into account the obvious things (UX, network effects, barrier to entry to use the product). All 3 of these are much better with Muse than they with are similar products.
UX is not a moat, especially in the era of AI software replication. I disagree that there are network effects for MUSE, maybe there will be at some point (e.g. vendors or something sign up for it) but I would love to hear why its better to be on muse when more people use it? Barrier to entry is somewhere they have an advantage, but not more of an advantage than google, or really anyone who has an app installed on your phone already.
definitely not me; I wouldn't doubt if Zster is working on turning muse into sleeper agents, _assuming_ chinese models will do the same, so they crib data and send it home like a good little maleware model.
In fact, we're going to see an explosion of the worst kind of AI model thanks to thinks mr Z thinks is in his best interest.
Remember, they had reports going back more than a decade about how they're harming children and destabilizing entire countries and as soon as Trumps in the whitehouse, they're licking boot.
If you’re falling for “sticky” interfaces, you’re not going make it. Your competitor won’t have the stickiness and the ability to use any harness and that will overcome Muse in the long-run. Let’s not forget, Muse is a close to SOTA model right now, but so was Google, and now they’re irrelevant. Google is one of many funded models that were close to SOTA, but we realized it’s benchmaxxed.
People are going to say that it doesn’t matter if you’re the best. If Muse, Grokbot, and Dot are able to reach mass appeal, it doesn’t matter if a person using a custom harness is better.
I respectfully disagree. There is a big difference between someone that’s running a custom harness that’s almost like a personal OS versus someone using something out of the box. And if you’re for using this to run your business, think for yourself who’s going to win, the person customizing everything or the one using what the big corps give you.
I’m talking from experience. I’m run a software factory building open source alternatives and Argon is the worst model. Wrecked multiple codebases. But every use case is different. If you’ve found success in yours, use it.
These are good points, though OP did specifically say that Meta is going to win "consumer" AI. I don't think the average consumer of AI is going to ever have to think about harnesses or worry about SOTA, they really will continue to interact with whatever is stickiest.
This exactly. The "harness" for consumer is not well defined but likely quite sticky if someone can build one. Meta is better positioned with its reach of whatsapp etc unlike anyone to try this.
Meta being meta, will likely run this as an "experiment mode" for 6 months to an year and if it does not work scrap it all and move onto the next thing.
There is a risk ofc muse does not work. But there is a good chance that with the ridiculous number of product teams doing A/B testing with insane number of people already on it, something will emerge that is interesting in a way it never could for their erstwhile Ar/VR foray.
The field is changing weekly at this point. Meta, OpenAI, and Anthropic have the most distribution right now so any can take this spot.
So why would you lock yourself in? They want you to lock yourself in, but if you build your own system, you won’t be. A layman will just use what’s best and get himself locked in, but they might be using an inferior model.
AI is a force multiplier. Doing the right thing with the right expertise is yielding dividends.
Many people don't even know what locking in means. The crowd in this forum < 1% of the world do, you and I do, and maybe a few outside. There are only laymen outside of 99%. No one i know of knows or uses a harness. They don't know the large number of VMs built on say Hermes. They really don't care or have time. If you give them a ready made product thats a click away, and its good enough, they will use it.
>For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software.
But this is what I’m saying, consumer AI is forever changing. Investing your time and money into a closed harness will put you behind someone running their own system that’s model agnostic.
If you’re on a forum like HN, would you realisticly invest your time and widgets into a closed harness instead of building your own? What if Meta’s model falls behind and all your things are on there? My advice is to invest in building your own because no one lab has shown they are worth using.
As far as “consumer,” I’m simply comparing consumer AI to a custom built one. One of OpenAI, Anthropic, Meta will probably win since they have the most distribution.
And, "stickiest", mind, where most are allready soul-sucked into the neverending doomscrolling dopamine nightmare that is Zuckbook - to beging with.-
First they "captured", or moated "communication". Now, they are going to capture "utility" ("I am actually doing something useful, not just doomscrolling. I can spend an extra 8h here. It feels good").-
> And if you’re for using this to run your business, think for yourself who’s going to win...
That's not consumer AI though. The whole "vs out of the box" is pretty much the definition of consumer AI. People don't want to fiddle with things. That's why UX matters. SOTA only matters in consumer if it translates to the UX. And no one has landed that yet. Many are working on it.
When I worked at Microsoft, I learned that the average user will fail to do ANY process that is longer than 2 steps. Let that sink in and then reread your comment. Nobody cares about SOTA outside of tech. They want the basic floor of “it does what I need at a reasonable price” that’s it.
I think any personal AI agent at this point is "very sticky", not just Muse. You really aren't going to switch when a lot of the value is in the agent memory and there's no easy way to export that, similarly to how you aren't going to switch away from Spotify when it's learned all your song preferences which you can't export. Right now, maybe the agent memory is small enough that if you ask the agent to export you'll get most of it out (Claude has notably made official export instructions to encourage switching from ChatGPT in the past), but in the future the memory will be too large and complex to do that effectively, and platforms could also block the memory export or degrade it deliberately.
So, in the long run you pretty much have to pick one personal agent and stick with it, and switching is costly (unless you are using an open agent that lets you just download the memory, in which case it actually becomes cheap, because AI agents make other parts of switching way easier in general). For the purposes of Meta this means whatever user they get they'll probably keep.
HOWEVER, the thing that Muse doesn't have is network effects. They're running into the exact same problem as ChatGPT in this area, IMO. There's no reason for you to be running specifically Muse even if all of your friends are running it. So that gives other vendors a chance to catch up.
You can export all your data from OpenAI and Anthropic. I also exported all of my Liked songs on Spotify and have a scheduled task to pull any new ones each day.
I don’t follow why “in the future the memory will be too large and complex to [export] effectively”.
to me, export should be easier and easier since you should just be able to ask the model for the memory files. it’s literally list text on a hard disk, even if there are some new formats in the future, they’ll just be files on a disk that another agent can then inspect and integrate. in fact, i think exporting data from services will become easier and easier since agents can do the work even if platforms try to block it.
> a lot of the value is in the agent memory and there's no easy way to export that,
Can't speak for muse and grok bot. But all other agents like openclaw, Hermes simply store then add files and it's trivial to migrate to another agent. Just point the agent to it
Bold statement. I wouldn’t count anyone as a winner on consumer market at this point.
I think the weight of people who dislike Meta is going to be a problem - that and forcing to give a lot of information just to use it is problematic for adoption.
Also honestly, the models don't have to get much smarter to handle the consumer side of things. They could innovate a ton on the harness to farm out requests and manage the state of the agent. They have an incentive to do this to reduce the infra usage so they can go wider with the agent.
I'm looking forward to seeing Muse devices like Amazon's devices but with superintelligence:
- Muse-equivalent of Alexa
- Muse-equivalent of Ring security devices: doorbells, cameras, alarms, etc
- Muse home automation devices (lights on/off, smart roller, smoke detector, water detecto, etc)
- Muse ring
...
Sure, just not nearly enough for sustainability I think. I'd reserve that call for the moment we can stop bathing in free VC money by using subscriptions.
Also, as far as my own use goes, I found that the capabilities/price ratio of DeepSeek's offering is very hard to beat. It's really good enough so the thing that matters to me (besides table stakes like uptime, latency, etc.) is price at this point.
Of course they are “profitable”, Dario said so last month (if you accept a definition of “profitability” under which every tech startup is profitable by the day they have their first paying customer).
They seem to be reporting gross margin by excluding the cut that cloud providers take for hosting their models, and excluding R&D.
Basically they seem to be taking credit for the whole amount that customers are paying cloud providers to run Anthropic models, rather than the actual net proceeds of that business to Anthropic.
> They seem to be reporting gross margin by excluding the cut that cloud providers take for hosting their models, and excluding R&D
The media is overstating what they're being leaked.
Anthropic is reporting traditional losses. It's also highlighting that operating figure, because it answers the question around what the business could look like in a steady-state and first-party-only mode.
They aren't claiming the profitable right now. But for an unprofitable company, having these sorts of metrics can give you a sense around when they will become profitable and what it will take. (Plenty of people were doubting that AI makes money on inference not too long ago.)
How about the fact that they're not shouting from the rooftops how they're managing to make money in an industry that is currently mostly known as a giant money fire? Instead they're reporting being profitable "after making certain adjustments", which sounds a lot like "not profitable".
> they're reporting being profitable "after making certain adjustments", which sounds a lot like "not profitable"
Adjusting out stock-based compensation and non-cash expenses is absolutely fine for a company growing at the rate Anthropic is. The concern is more that they cherry picked a couple months when their expenses were abnormally low [1].
> The rent market on intelligence never materialized
So where is the ~$100B combined revenue for OpenAI and Anthropic coming from? Open weights models have been close enough to the frontier for over a year now, probably longer, yet the $$ shows no sign of shifting.
OpenaAI and Anthropic models are priced quite competitively with open ones, they even cut prices by quite a bit since spring. So it's not too surprising that their revenue is pretty high, the question is whether the competition and downward price pressure from open model providers combined with their extremely high R&D spending is sustainable.
last year was google winning AI, this year is meta winning AI. I think what is true is Meta, Google will all find their place in the AI world and continue to make huge amount of money. A consumer will use many different AI products, each with a different use case. Everyone will make enough money to live well.
This is what stood out to me. They did the actual work of testing and documenting deployments on several platforms. It's a bit of upping the standard for open weights instead of just posting the blob and letting the community figure it out.
It makes sense from a provider standpoint. If you are going to give it away for free what is the end game? Helping build the world around your model such that it is a pathway that leads to greater adoption of your ecosystem is smart. Dev relations in a way.
Yes you need big compute to deploy this locally but $20-30K in hardware for specialty markets is nothing.
This isn't an accident. It is revealing a strategy.
Apple research has also posted information on the topic.
> In this post, we show how to run the latest frontier-class model — Muse Spark Open, with nearly a trillion parameters — right from your desk, using the latest version of MLX, released today. On a cluster of four Mac Studios, Muse Spark Open generates over 36 tokens per second
Beginning on a single Mac Studio (up to 512GB of unified memory), then scaling out across multiple Macs with the RDMA over Thunderbolt driver, below we share an overview of the range of models you can now run on Mac Studio.
day one support for all inference engines so that whichever kind of capable computer you have, you can run it.
you do need a really capable computer though
They suggest a pair of Mac Studios w/ M5 Ultra chips and 512 GB of RAM. A single Mac Studio w/ M5 Ultra and half that RAM (256 GB) is around $11k, and the 512 GB option isn't even available yet. So.... yeah.
Yes, but still, what about users looking for NPU support. What happened to Microsoft's copilot+ PC campaign? (the NPU hardware, not the logo key and ad campaign)
AMD's XDNA2 NPU is best supported by ONNX, I think Qualcomm Hexagon is too. The best consumer unified architectures are Apple, AMD ai+ and Qualcomm Snapdragon X Elite, even intel has an NPU offering.
NPUs are meant for tiny, tiny models that can run basically at all times but cannot affect battery life (e.g. translation, live-video cleanup, filesearch etc.), not for big LLMs. This is about 500 times to large for a NPU.
There are plenty of smaller less powerful open weights models if all you have to run it on is an NPU.
This is a much larger more powerful model, so running it locally will not be for everyone, but you can still run it via a hosting service such as Fireworks AI.
I guess that's the attraction - that they may be small enough (likely quantized) so that you can run them locally, so there is no privacy issues, and no-one can turn it off, etc, but technically "open weights" is just referring to models, of any size, where they give you the weights (basically give you the model) so you can run it yourself, instead of having to access it via paid API running on some computer that they control.
In light of recent pieces like "The Emperor Pelican Has No Clothes" (https://arcfu.com/post/the-emperor-pelican-has-no-clothes/), I'd be more curious to see how 'PelicanBench' becomes more robust in the time to come -- I feel like all models are now benchmax'd on the standard pelican SVG rendering task and we need something more multi-dimensional. It would be great to have a script that prompts a model to render the pelicans in several different situations (like the linked article) that showcase some kind of broad generalizability or other sort of 'transfer learning' going on with these models that would really then be impressive and worth noting on - otherwise PelicanBench is losing its relevance.
The gears etc on bikes are customarily on the right, so almost all photos of bikes (eg for sale, by hobbyists) face to the right to show off that, and so it makes sense for LLMs to have such a preference.
The model license requires large inference providers to contact Meta separately to obtain a license to serve the model. I know we've seen this before with other large models, but it's still a bit irritating.
The usual effect of separate licensing is decreased price competition from public providers (like we saw with Kimi K3). I guess it's better for the providers if the price is fixed: any efficiency improvements directly improve margin rather than driving prices down over time. Still, to provide my own anecdata, the main thing keeping me on a paid subscription is token costs for the more capable models.
Then it's not open-weight. Open-weight means "weights available under an open-source license", and licenses prohibiting commercial use are not open-source.
The nearest analogue would be "weights available", I guess.
There's obvs decreased price competition, but it also does help with ensuring model quality.
Quite a lot of providers on OpenRouter run degraded or broken models. I think some is deliberate, but I think a lot is just providers not setting up models properly; especially with the rush to day zero models.
I have just tried to download the models, but it is not that simple, because they are "gated models".
This means that you must first create a personal account at Huggingface.co, then you must fill a request form for Meta, including whom you work for and in which position, submit the request to Meta and wait for it to be approved.
In my case the approval was received in about 10 minutes or so.
Presumably after that you are able to download the files, though I will have to see how easy it is to use your credentials from a CLI download program, because downloading a big LLM with a browser is completely impractical, as the transfer of the big files breaks very frequently so you need a download program like wget or curl, which will restart the interrupted transfers as often as necessary.
This is kind of a standard at this point for many models (including prior meta models like llama)
You create a hugging face token, set an env variable (HF_TOKEN) and download it with the hugging face CLI or just automatically as needed with trandormers, vllm, etc…
It seems like meta and mistral are both hedging their "best" claims geographically because they've realized they don't have to even beat deep seek anymore, they can just win customers by being "the best one I'm allowed to use".
re: "Apple’s unified memory architecture lets the M5 Ultra's integrated GPU use up to 512 GB per Mac Studio, so two machines can hold an entire model. And the chip’s 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs. The result is two Mac Studios behave like one supercomputer with a terabyte of unified memory."
At the current market price of Apple Mac Studio Ultra, 30-core CPU + 64-core GPU w/ 256GB RAM is already reaching at around $10k. Keep in mind that computers are appreciating assets recently.
The 512GB RAM option is not even for sales yet (they said late October availability). Based on typical Apple pricing, it's likely to be around $14k apiece. So we are looking at least $28k - $30k for the setup Muse team is advertising. That's essentially the Nvidia H100 GPU w/ 96GB VRAM/HBM3 at up to 3.94 TB/second memory bandwidth, 3x+ than Mac Studio's Unified Memory speed.
Also, how did Meta get access to Apple's unreleased M5 Ultra w/512GB RAM Mac Studio? Or is that simply a marketing piece that they wrote purely on specs?
> Also, how did Meta get access to Apple's unreleased M5 Ultra w/512GB RAM Mac Studio? Or is that simply a marketing piece that they wrote purely on specs?
It should be well known that Apple provides early access to developers under strict conditions. Otherwise, how do you think they announce new hardware not for sale, and show demos of third party apps taking advantage of such hardware?
The 512GB is obviously geared towards local LLM crowds. I'm just speculating, but I would be surprised if there's not at least a few devices with the 'ecosystem players' like LM Studio, Unsloth, etc. Of course, they'd never confirm nor deny :)
Also you are comparing apples to or— dang it, I don’t mean to do that in this thread. Well, it’s a very poor comparison, as they are completely different computing devices with completely different use cases, capabilities, and customer bases. Not to mention power requirements!
I love this. He's used up all the Thunderbolts on the back to make use of the new hack where you can connect all of them, so he has to plug his monitor in on the front :D
OG Deepseek Flash was pretty mediocre and then got a lot better with post training. The other thing they are fighting is most/all of the chinese labs are free to use/train on the 10B+ SOTA reasoning tokens they "stole" over the summer March-July 2026, if meta is playing by the rules, they are building this from the ground up. Getting anywhere near Deepseek Flash without that free training data is pretty impressive feat.
OG DeepSeek V4 Flash and V4.1 Flash are actually completely different architectures. To wit, V4 runs fine on 2x96GB cards but you’re not running V4.1 unquanted on less than 4x96GB.
It’s not really an apples-to-apples comparison.
(I’m not sure why they didn’t call it V5 Flash, but model names from most labs are completely nonsensical and disjoint from their underlying architecture.)
I partly agree with the sentiment, but this is also the first time I've seen an actual lab announce they optimized for consumer hardware (very expensive, but still consumer).
It's kind of absurd that no non-quantized MOE model exists that's sweet spotted for 128GB Macs, for example. I hope more open labs follow this lead but also offer lower tier models, perhaps specialized on certain types of tasks
How does this compare to Muse Spark 1.3 contribute? I've used Spark 1.3 a lot lately , since its currently free to use via OpenCode. I haven't had the best experience in the past with models from Meta. But Spark 1.3 is has been surprisingly great. Especially for tool calling.
Overall it actually does pretty solid. It doesn't completely follow the design, but it gets the vibe. It's one of the best open models I've seen for conversion, which is impressive.
And disclaimer since this comes up with these - objective with this test is to be a good image->html test, not a good web design. Also it's meant to use 3js to render the planets, which if you don't have GPU acceleration enabled for your browser will lower the FPS significantly.
This is great progress for "open-weight" models. Quotes due to the restrictions, but still progress none-the-less. Looking forward to seeing any quants that all mere mortals to run this at home without early access Apple hardware.
Well, this is much better than nothing -- we should be grateful that whatever the market conditions and corporate incentives, a major lab is releasing a capable open-weight model. But still very far behind DeepSeek and GLM. Even on consumer hardware, Qwen 3.8 27B is going to compare favorably to this.
True, but the conditions are quite relaxed, so it can even be used commercially without worries by small businesses, while it is unconditionally free for non-commercial uses.
Other big open-weights LLMs have similar licensing terms and I believe that they are quite fair, i.e. at the point when someone uses such an LLM to obtain tens of million $ of revenue it becomes fair to demand a share of that for the creator of the model.
Its interesting that muse benchmarks are limited to open-weight models. while when glm introduced 5.3 and its flash variant they were compared to kimi, Anthropic and Open ai.
I know they say that there the biggest open-weights model in us yet but 1.3 was decent but not as useful as Chinese variants like glm, deepseek and mimo.
I think a lot of people in this thread are openly dismissive: "oh its just open claw, anyone can and many have build that".
I think what they are missing is the fact that there are not that many teams in the world that can deploy muse like product to A/B test a lot of hypothesis at the scale meta can at scale. I mean even if you could get to scale with DC you need a reason for people to use your product.
My mom and dad and their family have never heard of Claude and don't use chat gpt except once in a while as a search engine. But meta will figure out a way to insert this into modals and feeds to make people stumble and stick on it.
I am no meta fan, but I think people underestimate how fast product uptake there is for non Dev harness.
It's not just openclaw, there's OpenAI's Dots and Gemini Spark.
Facebook doesn't have an real advantage to drive adoption other the being free. Which is not a moat because Google, certainly, and OpenAI, very probably, would offer this for free, in the event this capability turns out to be important to consumers.
Gmail, which half of the people in the US use their daily e-mail is certainly strong thing to build an agent on top of.
Just because meta can shove stuff in the faces of people doesn't mean they can succeed at any random thing. Most Facebook users distrust it, and dislike Zuck personally (at least that's what his lawyer said in court) and that's a rather large barrier to succeeding with agents, where trust is foundational.
But my assumption about the speed of product uptake may be skewed by not-valuing Meta's "consumer task running" pitch. But if people actually want this, for some reason, competitors at scale seem very likely to appear instantly.
This really seems more about meta AI team needs more users of their AI to do a good job building AI, and doing this for free is niche where they can get a few more users, rather than this being genius gambit.
Coding was not the main application before ~Dec 2023, the same way the light bulb was not the main use of electricity for a long time. Meta for better or for worst is good at discovering new use-cases for their products(however bad they are for society is a different debate) and pivot if they are not - remember the Portal device( https://www.meta.com/portal/ ) ? Good hardware, decent usecase, killed within a ~year of its release.
Zuck said by himself that they will release the weights of Muse Spark 1.2, and now we only get that one. At least they bench similarly bad for the size on Artificial Analysis
In terms of performance it seems like this has somewhat worse benchmarks than Muse Spark 1.3, closer to 1.2 but not the same. Of course significantly worse than Muse Wonder Full.
The exciting thing about this is in them optimizing for inference on Macs. This, and optimization for other consumer hardware, is where local AI needs to go
Given the size of Macs (2 or 4 of them) required to run it, they're just paying lip service to that option. Other models, e.g. deepseek flash, are much more practical size. Even that takes a 96 or 128GB ram Mac.
It's really nice that all AI pages are using canvas animations. A few days ago, Gemma 4's animation wasn't that impressive. Same with Muse. So far, the best animations I've seen are from OpenAI and Anthropic.
Ehm.. I have a model that runs on my older gaming desktop that scores as well (or better) as this model does on DeepSWE. This model does better than Flash Next in many areas and I am sure is much stronger overall, but hmm.
Who would have thought that one of the top American tech companies would boast about how their latest product is the best in class in the US, not the world, because they're behind China.
Very big LLMs can still be useful even when you run them only at the speed of around 1 token per second that can be reached with almost any mini-PC or laptop with the weights streamed from SSDs.
While at that speed they will not be your main LLM, there are tasks that are still useful even when requiring many hours, e.g. an overnight run, for instance for checking what has been produced after many interactive iterations of a fast but small model.
Or rather capturing the people Anthropic keeps alienating.
The Zuck has (in a different context) said that a few billions wasted are a decent bet (or something along those lines), so I think they're playing extra long games here with no real focus on profit any time soon.
It performs worse than Muse Spark 1.3 and 1.2. And it has an outdated attention mechanism which at 1M tokens will need over 200GB just for the KV cache.
Very disappointing release. We're lucky to have Chinese models.
great, I've been reading series of papers related to open source/open weight LLM Backdoors.
It's 100% possible, curious when will be the first incident.
Having something built by entity with HQ in the wester world is always a bit safer. That could be mistral, but they decided to be a consultancy company, not an ai lab.
everyone is doing backdoors, except eu, they just receive bribes from those who install backdoors.
the only difference is: the backdoor is made to kill people like you or not. I'd rather have a backdoor that is installed to kill someone on the different continent, e.g Iran leadrship, not in my country/city.
This post could be framed with golden filigree, titled "The Most HN Comment Ever Written" and perhaps displayed near other artworks like a Juicero, Flock camera or Sam Altman's disinterred belongings.
Orange Site will have a properly interesting history museum, at the rate we collect curiosities like this.
I don't think this is going to change anything for most people, but it certainly fits a niche for US-based companies that refuse to run Chinese models and only want to run their own inference.
But I'm really happy to see Meta do this and I hope they continue to release open models in the future. Open models are really the only way to ensure fair access to AI for everyone.
I do not know about you, but I will probably also run this model, even with a mini-PC like an ASUS NUC, albeit very slowly.
Even if for your main LLM you want a speed high enough for interactive use and fast iterations, it is still useful to also run a much bigger LLM from time to time, even if that requires many hours for completing a task, due to the slow inference speed when the weights are read from SSDs.
There are tasks where the quality of the result is more important than the time needed to produce it. For example you might run all day a great number of iterations of a fast small LLM, and then run overnight a big LLM to check the end results of that work.
Therefore we need both small open-weights LLMs that can be run from the GPU RAM or at most from the CPU RAM, but also the biggest open-weights LLMs, even if the latter can be run only from SSDs by most people.
Not at an appreciable token rate.
I wonder if these large open weight models are considered safer to release for that reason?
I'm thinking no because any small to medium size criminal organization or state funded group will be able to cobble together the resources to achieve some kind of runtime.
Even at 10 tok/sec, making bioweapons at 10 tok/sec is concerning.
You do realize that a high school Chemistry course actually contains all the details you need on both bioweapons used in WW2, right? Plus several others you could use. Those books contain the answer to why those 2 people died in the RV in Breaking bad. And they should. They also contain information on Mustard gas and Zyklon-B, including preparation instructions, but of course, you'll have to puzzle it together. That these substances can be used as bioweapons is heavily deemphasized, but you just can't reasonably avoid mentioning and showing them.
Also these books contain maybe not the solution to making thermobaric explosions, but certainly the principle and show you exactly what problem you'd have to solve to make those weapons.
Just about the only thing they stopped teaching is that if you combine that knowledge with which plants contain what ... there's an incredibly easy way to make pudding with cyanide that's safe to touch, for example. That they suppressed.
But at least for Europe, the number of lethal plants just growing along essentially every road or in every forest numbers in the dozens. Including a very famous one they teach every kid about.
Of course, in both cases, the real challenge is making them and deploying them without killing yourself.
We are not safe from bioweapons because nobody knows how to make them. That's not it.
It seems like you're talking about chemical weapons.
Which is fair, they are very easy to make without the assistance of an LLM, and I don't hear about that kind of thing in the news.
I was talking about biological weapons like gain of function and that kind of thing.
The implied question was 'Should companies like Meta publish open weight models when they could theoretically be subverted by malicious users?'. So far as I'm aware there are few people worrying about this kind of thing. Bill Gates for one. But apparently it's off topic and/or not interesting, or not welcomed as a discussion point.
What is the point of releasing "The Most Capable US Open‑Weight Model?" I can understand why I might care about the location of a model serving provider, but why should I as a model user care about the location that an open weights model was trained?
It appears to already be far beyond that, they are working with each other[0]:
>ChatGPT, Claude, Grok, DeepSeek, Perplexity, Gemini, Copilot, Le Chat de Mistral and Meta AI filter your data to Google, Meta, and several of the main data brokers.
Only if NVIDIA didn't shit on consumers, maybe we could run this with a single card today. But they are not even selling the 5090 32gb anymore. I hate this progress. Models are out, but there's no hardware to use them.
”Serve the model on either two Mac Studios, with M5 Ultra and 512GB of unified memory; or four Mac Studios with 256GB of unified memory, for a total of 1TB of unified memory, enough to hold a NVFP4 quantized version of Muse Spark Open”
Closer to ~$30K if I had to guess. A 256 version is $11,299.00 (2TB SSD, 256GB RAM, 36-core CPU, 80‑core GPU), so yeah, 4 of those would be close to $50K but the 96->256 RAM upgrade is $4K, if we assume 512 will be double that it's ~$15K/ea.
But yes, still a massive amount of money and that's before we know the TPS or any other stats.
How much does the engineers at Meta make? Spend enough time together with others making $250k+ and eventually you start suggesting that two 512 GB M5s is something you can put on your desktop. I mean, it's one banana, Michael. What could it cost? 10 dollars?
I tried Muse Spark 1.3 for a while because it's so cheap, but it's definitely in a lower league than Astra, Opus 5.5 and Argon. Still produces slop if you try and vibe code with it, whereas the more capable models write good code.
Tech people really are drug addicts. Rant about Zuck and Meta for 364 days a year and then jump over each other to be first in line the moment he shows something shiny.
Nemotron, Inkling, Gemma, Trinity Large, Snowflake Arctic were all released/updated in the last 6 months, and there are countless other smaller models out there. US open weight is very far from "dead", and certainly doesn't need Zuck as a savior.
As an aside, it's really unfortunate that artificial analysis only provides comparison at the highest reasoning level, which is mostly unusable for high-volume production use cases. I would really like them to start providing comparisons at low and medium reasoning levels across these models because those tend to be the ones that actually get used in production.