This just made any closed LLM a huge supply chain risk. Everybody was aware of this possibility, but now it actually happened. It's like having nuclear weapons vs. firing a nuclear weapon.
Especially outside the US customers are going to be very hesitant to keep adopting LLMs from US companies.
> Especially outside the US customers are going to be very hesitant to keep adopting LLMs from US companies.
Not really. There aren't any other choices, and the PRC also heavily utilizes export controls [0].
This is why sovereign AI has become important, as can be seen with EU NatSec uses cases tending to use Mistral [1] and Indian governments starting to use Sarvam [2].
That said, for most commercial usecases, older generations of Opus as well as enterprise grade GPT and Gemini are fairly good.
The distilled OSS models are alright for hobbyists but if you have actually used unrestricted and enterprise grade versions of Claude, Mythos, GPT, and Gemini (most hobbyists don't get access to these) you see how far behind the open weight models are.
Even in China, traditionally open minded models teams like Alibaba's Qwen are looking to become more restricted given the org changes [3].
Also, Corporate RFCs now demand final say on model used and depending on the geo, this can be a dealbreaker (eg. An American financial institution will absolutely blacklist a vendor if they use a Chinese model and same in reverse and European defense vendors mandate sovereign EU models depending on the opportunity).
> if you have actually used unrestricted and enterprise grade versions of Claude, Mythos, GPT, and Gemini you see how far behind the open weight models are.
I really do feel like DeepSeek V4 Pro is often better than current Sonnet is, in the general case.
Opus 4.7 is a solid step above Sonnet, and Fable was a solid step above Opus 4.7. I've only had Fable for a few days, obviously, but I was decently impressed after Opus 4.8 being a downright disappointment for me (it's just too buggy; I had it go out of control 3 separate times on things Opus 4.7 never had any trouble with.) I still ran into limitations. It's not world-endingly great.
So, based on that, I think DeepSeek V4 Pro is, ignoring multi-modal capabilities, about a couple solid steps behind. Assuming model iteration will continue to decelerate, especially as Anthropic heads into IPO, I'm guessing that DeepSeek will probably be able to strike back with something further along. Of course we'll see how able and willing they are to stay open weight, but they've done well so far so, no reason to doubt them at the moment.
(There are some models that claim to be ahead of DeepSeek V4 Pro. I've tried some of them and really not been that impressed. Maybe it's a me issue.)
Now I reckon that most people just simply don't really need Mythos/Fable for most of what they do and using Mythos/Fable tokens in place of Sonnet-tier models would not make any sense. At my job we already mostly just use Sonnet as it is. I'm sure there is some cutting-edge research where you want the absolute best model available and sure, in that case, you're stuck with Anthropic for the moment.
But is that really everyone? After all, while Mythos was dominating the hype cycles, quite a lot of impressive LLM-assisted CVEs dropped that were not linked to Mythos.
you think its that hard to get trade secrets from some openai or anthropic engineer if you promise them anonymity and a new better paid position? hell they might even give it away for free if they think what their company is doing is morally wrong. know how is not source code, you cant catch it with dlp or online leak scanners. you would need 24/7 combined human and electronic surveillance and thats something even the cia reserves for top level targets, it takes too much manpower to use it on everyone.
SMIC hired hundreds of TSMC employees and now its a couple years away from 3nm equivalent chips in full production. export controls only work against poor countries with less advanced industry like russia. china has the resources and export controls give the motivation. and if the eu/us relations get even worse i wouldnt be surprised if the dutch government let asml start selling euv machines to everyone just to get back at trump.
Compute was constrained. There is a lot happening, especially with chinese chips which currently points to a massive upcoming increase in non-US capacity.
Also, the EU, Japan, SK, ASEAN, and India are not supportive of using Chinese tech after China export controlled rare earth exports last years [0].
Software supply chain regulations also make utilizing Chinese software risky for ExChina players and make using ExChina tech risky for Chinese players.
Expect to see RFCs now demanding visibility into what models are used and right of refusal - this is already the norm in F1000s. Similar ones are likely to arise in the EU as well with some of the upcoming industrial policy changes being proposed.
If you’re talking about TSMC Arizona they aren’t fabricating at N3 until end of next year at the earliest, N2 isn’t slated until “end of decade”. I think they’re manufacturing Blackwell there which is N4 / 4nm
They've already been labeled a "supply chain risk". Probably not a good idea to upset the regulators more. Maybe tomorrow Opus will be declared too dangerous for the public.
You're mistaken, this is a cratering of the userbase inside and outside of the US. The ban is on any foreigner whether abroad or living in the USA, so Anthropic has no choice but to completely shut down access to the model for the whole world including the US.
Their IPO is well and truly fucked now. This also means no other frontier lab in the US is allowed to exceed Opus 4.8 capabilities.
If you're a luddite or a decel you should literally be dancing in the streets right now. And, if you're a tankie you'll be dancing right next to them. And, if you were hoping for a Star Trek-like future, you just adjusted your timeline for the worse.
>this is a cratering of the userbase inside and outside of the US.
Is it really? It was limited release anyway (like hypebeast merch!). Everything people are gonna talk about for a week is gonna be about how Fable was so cool that it got banned by the feds. If it's just the Trump admin being the Trump admin, Amodei is just gonna have to pay up as a racket / marketing expense. Or it is like I'm suspecting and this was pre-bribed and the ban is kabuki theater.
>And, if you were hoping for a Star Trek-like future, you just adjusted your timeline for the worse.
The funny thing is that solar and batteries advancements are actually this, not LLMs, but your framing kinda fits anyways.
The main error of the AI bubble is expressed in the The Jetsons cartoon from the early 60s.
In the future, everyone obviously would be running nuclear powered cars. It was just an engineering problem to be solved. Ford made the Ford Nucleon prototype in 1958.
The nuclear optimism completely blinded people to the ridiculous idea of an individual handling nuclear material for personal use.
The AI bubble error is this idea that everyone is going to have "AGI" in their pocket. It is just a completely absurd idea that is not going to happen.
Fable was interesting from what I tried but nothing close to AGI yet here we are. The models don't get smarter and LESS restricted from here.
To me, right away it seemed that the "Mythos moment" was extraordinarily bearish for the assumptions the AI bubble is built on.
Is it now? From a company's point of view, does it really matter that some expensive tool is allegedly good or not if it's reliability/availability is poor and subject to completely arbitrary and unpredictable change?
And they don't have to actually serve expensive model compute and this all goes away once they contribute to the right charitable organizations and patriotic causes funneling money to the right people.
This is quite clearly corporate capture of the white house by a competitor influencing policy, but it's hard to imagine something that plays more into anthropic's hand. They now own the model that was so good the US government made them shut it off.
it may be really good pr, but it's really bad for their IPO. If their market for future models is usa only, their potential revenue is cut by 50% at least. (and it's even worse because it means Europe, India, and China will all have companies making their own models that anthropic needs to stay ahead of)
Another sibling thread already called this out, but mentioning here: it's not "USA only", it's "US citizens only" (and I'm not entirely sure how dual-citizenship interacts with this, but I assume you can't sell to them, either, since they are by definition also foreign nationals). A private company only being able to do business with folks they can verify are solely US citizens (who themselves are also willing to submit verification of said citizenship to a private company), has a relatively small pool of potential users.
And so if this policy holds, Anthropic has functionally had Fable killed by government intervention, and in a logically consistent world, this would imply all other US-based AI labs may also never exceed existing (read: Opus) capabilities.
Regarding the dual-citizenship, you are wrong to assume that. To US government you are US citizen and that is all that matters, even if you have 5 different citizenships government and justice system don't care, you need to follow the US laws and can't cherry pick what you want. Regarding users, for any of this big 3 (Alphabet, Anthropic, OpenAI) only important customers are enterprises, not individual users.
I don't think so, most of enterprise customers are US based companies. They will basically give Mythos to US citizens in R&D while others will use Opus. I hope this is not the actual intent.
How many entreprise customers that aren't in the defense sector currently have R&D departments entirely composed of US citizens?
And what does it mean for indirect access to the models, through say agents working off ticket systems.
The problem here is that the valuations of these AI companies was based on the fact that they'd keep improving models. A company that just serves the latest Opus isn't worth trillions.
You think Anthropic will ask all their enterprise customers to provide passports for all their employees and then setup individual Claude accounts for each and every employee to gatekeep access to Mythos? Because a plain ole api key no longer cuts it
