{"slug": "meta-s-open-weight-whiplash", "title": "Meta's Open Weight Whiplash", "summary": "Meta Platforms Inc. is shifting back to releasing open-weight AI models, a reversal from its recent closed approach, as competitive pressure from China's cheaper models and rivals Anthropic and OpenAI reshapes the AI landscape. The company, which had previously paused open releases after hiring Scale AI's team to fix its Llama models, now argues it never abandoned openness, though critics note the timing coincides with Meta's lagging position. This move comes as Chinese open-weight models gain traction in the U.S. despite geopolitical tensions, raising questions about the true nature of 'open' AI.", "body_md": "# Meta's Open Weight Whiplash\n\nI'm old enough to remember the last time Mark Zuckerberg was [embracing \"open\"](https://spyglass.org/zuckerberg-llama-open-source-ai/) as [the future of AI](https://spyglass.org/zuckerberg-ai/). Which is [to say](https://spyglass.org/meta-llama-ai-zuckerberg-open-ai/), I'm at least [two years old](https://spyglass.org/meta-open-ai/).\n\nA lot has changed since the last time Meta released an open weight model – namely, seemingly *the entire team* building models for Meta. Famously, [\"Llama\" was put out to pasture](https://spyglass.org/open-source-ai-was-the-path-forward/) as Alexandr Wang and his Scale team [was brought in](https://spyglass.org/meta-scale-ai-reset/) to fix [what ailed those previous models](https://spyglass.org/metas-moe-ai-mistake/). That included not only [spending billions](https://spyglass.org/hackquisitions-hackquihires/) on [fresh AI talent](https://spyglass.org/zucks-eleven/) and [data centers](https://spyglass.org/metas-ai-relativity-theory/), but also [setting aside](https://spyglass.org/metas-open-source-ai-mistake/) that notion of \"open\" in favor of a closed approach to AI development. Why? Because it sure seemed like that model – and those *models* – had won the day just over a year ago.\n\nTo hear Meta tell it – as I'm sure we're about to *endlessly* – the company never left \"open\" behind, they simply did what they needed to do in order to reset their efforts and catch up, fast. And there's undoubtedly some truth to that, but also some gaslighting. If OpenAI and Anthropic were still running away with the AI game – or if Meta was in the lead – we probably wouldn't be hearing too much about \"open\" right now. But again, a lot has changed in the past year.\n\nMost notably, OpenAI slipped and fell behind their chief rival Anthropic. And Anthropic, in turn, has kept trying to commit seppuku to [sacrifice themselves](https://spyglass.org/amodei-another-day/) at the hands of [an angry Trump administration](https://spyglass.org/anthropic-vs-administration/). This, mixed with their new frontrunner status has led to an [inevitable backlash](https://spyglass.org/anthropic-becomes-the-villain/) against Anthropic, even though they mostly maintain their lead in models and in the all-important coding category.\n\nBut the real key was China seemingly catching up to that frontier of AI. Not quite, but perhaps close enough – [closer than](https://spyglass.org/deepseek-is-seeking-you/) DeepSeek 18 months ago, and with AI in a more mature state, being \"close enough\" perhaps matters more now. And that's especially true given the [cost and consumption concerns](https://spyglass.org/ai-overkill/) that a lot of corporate America is feeling at the moment with regard to AI. If those Chinese AI models are \"close enough\" at a fraction of the cost of the frontier models...\n\nAnd while it may sound crazy to think that American companies would willingly use Chinese AI models in a time when the current administration has made it clear that China is the US's techno enemy #1, China left a door open to blur such lines with the focus on yes, open weight models.\n\nIf anyone can download these models and host them elsewhere – including on the clouds of the Big Tech players in the US – what's the problem? I mean, there *are* potential problems in that no one knows exactly what went into these models to train them as they're not actually \"open source\", as that's not the same thing as \"open weight\" despite [what op-eds](https://www.washingtonpost.com/opinions/2026/07/20/open-model-ai-is-good-competition-anthropic-openai/?ref=spyglass.org) and [ New York Times headlines](https://www.nytimes.com/2026/08/10/technology/meta-ai-open-source.html?ref=spyglass.org) (see below) would have you believe. Still, much as was the case with the \"DeepSeek Moment\", I'm not sure how much the Chinese models will\n\n[end up mattering](https://spyglass.org/deepseek-moment/)– as they're already\n\n[tweaking](https://venturebeat.com/technology/kimi-k3s-full-weights-are-here-but-theyre-open-with-a-caveat-what-enterprises-should-know?ref=spyglass.org)\n\n[their terms](https://www.reuters.com/business/retail-consumer/alibaba-plans-charge-big-users-its-next-open-source-ai-model-sources-say-2026-08-07/?ref=spyglass.org)\n\n[and models](https://www.bloomberg.com/news/articles/2026-08-06/deepseek-plans-significant-price-increase-for-its-ai-services?ref=spyglass.org)in showcasing\n\n[why \"open\" here](https://x.com/mgsiegler/status/2085661729409716433?ref=spyglass.org)very much deserves the quotes – versus what such models\n\n*showcase*.\n\nThat is: open weight models may now be a viable alternative to the closed variety and, relatedly, distillation of such models may actually *spur* innovation in the sector.\n\nThat's why we saw hundreds of US companies sign on [to a letter](https://spyglass.org/open-weight-ai-models/) to help ensure that the US government [didn't destroy](https://spyglass.org/open-vs-closed-ai/) this \"open\" path in their attempts to slow down China. Obviously, the initial batch of companies that signed on, from NVIDIA on down, were conflicted with such ideals. But it created the intended groundswell to seemingly both get the government to heel but also pressure current \"closed\" AI leads like OpenAI and Google to sign on.\n\nYes, yes, both offer \"open\" models too, but only far less powerful varieties, [released months after](https://spyglass.org/openai-open-weight-model/) their frontier models. That is not what China is doing. And that is\n\n**not** what Anthropic wants to become the norm. (As they still refuse to sign the letter despite the pressure because\n\n[they say](https://www.anthropic.com/news/position-open-weights-models?ref=spyglass.org)while they're generally okay with open models to some extent,\n\n[they're worried about](https://spyglass.org/ai-nuclear-weapons/)the\n\n[safety and security implications and ramifications](https://spyglass.org/claude-mythos-quandry/)if the frontier goes open weight. Which many read as the company protecting their moat. But others note that this is likely\n\n[what the company actually believes](https://spyglass.org/what-if-anthropic-is-right/)from Dario Amodei on down.)\n\nAnyway, that's a long-winded way – though about 1/10th as long [as Mark Zuckerberg's new \"essay\"](https://www.meta.com/thefutureisforeveryone/?ref=spyglass.org) on the topic – of saying that Meta is back in the \"open\" game. Again, a game they'll say they never left, but [they really did](https://spyglass.org/hackquisitions-hackquihires/). And that actually may have been a mistake, [as I noted a few weeks back](https://spyglass.org/open-weight-ai-models/). But it's also probably not too late to correct as the US \"open\" model race still feels wide open. And without Anthropic, OpenAI, or even Google fully committing to it (again, at the frontier), Meta may have an opening here.\n\nAnd they're taking it.\n\nTo be clear, 'Muse Glimmer' is not at the frontier. Hell, it's not even at the frontier of Meta's own current model offerings – meaning, Muse Spark 1.2, which is by most accounts a very good, [but not frontier-level model](https://spyglass.org/meta-muse-spark-ai/). They note that Muse Spark 1.2 will also get its weights released at some point soon, but again, Glimmer is not that. Which suggests that, [as expected](https://spyglass.org/meta-open-ish-ai-models/), Meta is going to follow the new general \"open\" playbook, at least in the US. That is, release a model, and once it's out there in the wild for a while, then you can \"open\" it up.\n\nWhy they didn't just wait to do this with Muse Spark if it really is that close to being \"open\" too, I don't know. But it seems like [they aim](https://www.bloomberg.com/news/articles/2026-08-10/meta-releases-muse-glimmer-ai-model-people-can-run-on-their-laptop?ref=spyglass.org) for Glimmer to be smaller, to the point where it can run locally on some machines.\n\nThe real key will be what happens when Meta releases their *actual* frontier model, the one codenamed 'Watermelon', which many expect to be both coming soon and also competitive with said frontier. Will that too be \"opened\" up at some point shortly after launch? Meta isn't saying, but also, to be fair, they're not saying anything about 'Watermelon' beyond nebulously alluding to it. Because it's not actually out there yet. And talking about models before they're ready has gotten Meta in trouble before – see: Llama 'Behemoth' which no one ever did end up actually seeing in the wild.\n\nMy guess would be that if 'Watermelon' really is as good as the current frontier, the weights won't be released but it will instead be used to *distill* some other Muse variant that Meta will release as an \"open\" model... We'll see!\n\nRegardless, it's clear that Meta, and Zuck in particular, are slamming on the gas. That's why you get a 7,000-word essay just a couple weeks after writing a decidedly more parseable version as [an op-ed in The Wall Street Journal](https://www.wsj.com/opinion/the-ai-future-is-for-everyone-a0c24e20?mod=article_inline&ref=spyglass.org). This memo is verbatim to that at many points. So this is like the \"Director's Cut\" I guess. Boy does it need an editor...\n\nMeta clearly sees \"open\" as a new opening here, which, yes, is ironic given that it was their initial path. But they also see openings in consumer/personal AI, especially with OpenAI [seemingly taking their eye off that ball](https://spyglass.org/theres-always-a-siren/) in their fight with Anthropic. As well as an opening [on the business model front](https://spyglass.org/you-are-the-tokens/), where [margins are always opportunities](https://spyglass.org/meta-ai-cloud-business-model/). Still unclear: can Meta make any of this pay off [this time](https://spyglass.org/llama-milk/)? They seemingly have [more options now](https://spyglass.org/meta-cloud/), but this is all completely unproven for them...\n\nZuck also clearly sees an opening around [the messaging of AI](https://spyglass.org/ai-steve-jobs/). He wants Meta to hold the \"positive AI\" position. That's going to be easier\n\n[said than done](https://spyglass.org/meta-many-bets/), quite literally. But yes, \"open\" is a part of that messaging. So \"open\" is back! No llamas this time though, sadly.", "url": "https://wpnews.pro/news/meta-s-open-weight-whiplash", "canonical_source": "https://spyglass.org/meta-open-ai-muse-glimmer/", "published_at": "2026-08-10 19:55:39+00:00", "updated_at": "2026-08-10 20:12:19.014894+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "generative-ai", "ai-ethics"], "entities": ["Meta Platforms Inc.", "Mark Zuckerberg", "Scale AI", "Alexandr Wang", "OpenAI", "Anthropic", "DeepSeek", "Trump administration"], "alternates": {"html": "https://wpnews.pro/news/meta-s-open-weight-whiplash", "markdown": "https://wpnews.pro/news/meta-s-open-weight-whiplash.md", "text": "https://wpnews.pro/news/meta-s-open-weight-whiplash.txt", "jsonld": "https://wpnews.pro/news/meta-s-open-weight-whiplash.jsonld"}}