villish

Have you tested older Gemini models? I wonder if their anti-distillation security measures are silently downgrading you to a less capable model.

> Spitballing random pr pieces from companies about to IPO is not helpful to anyone, fyi.

You can just use logic to come to the same conclusion. The Chinese labs have a fraction of compute and data. The simplest explanation is distillation.

Muse is significantly faster and more token efficient. That is pretty impressive considering how much larger it is. There is a lot of potential even though this release won't shake up the standings.

China blocks ChatGPT. Which is “banning the competition” that the above posters were claiming the US would do.

As far as I’m aware, only China does this currently.

You should not trust any LLM to give you accurate legal advice. I can’t imagine all the ways that could go wrong.

They are more focused on incorporating AI into their products, for better or for worse, but they could burn a whole lot of compute to train a model that competes for the #1 spot. Is that worth it to say you’re the (temporary) leader in AI? I doubt it.

Do you use Mistral's models to help improve the European offering?

Otherwise if you don't care about your data being used in training you might as well use the best, which is the US models on a subscription plan. If you do care, use downloadable weight models served by reputable providers, or selfhost.

Potentially higher cache hit rate at the expense of your data being kept for training.

The European labs are distinctly disadvantaged. US labs scraped the entire internet and downloaded copyrighted works to train their models and just paid a fine. They get 99% of all investment money in AI because of the lead they hold. Chinese labs also scrape the internet, download copyrighted works, and launch massive distillation campaigns using stolen accounts and credit cards. They get full support and funding from the PRC.

Europe gets barely any funding because it is so far behind and cannot use the illegal strategies Chinese labs use to distill Opus/Astra. Mistral cannot compete because the competition is literally breaking every law.

> What's your point again

I said it in my original comment. That is the moat. The reason frontier AI is a two horse race. European labs cannot gain ground because the only way to do it is illegally.

> Anthropic would be running their own distillation “attacks” too if _they_ were the ones playing catch-up

Probably. But if OpenAI or Anthropic stole your credit card to purchase tokens you could sue them. You won't get a cent from any Chinese labs.

> it’s nice to see some actual competition again.

Competition benefits everyone. But this isn't fair competition. A German startup cannot legally do any of these tactics required to bypass Anthropic/OAI's counter-measures. Which makes EU less competitive and therefore less investment in European AI.

Yes but Claude is filtered by the Great Firewall. Anthropic also restricts Chinese access. Their security measures aren't bulletproof but they do catch a substantial amount of those accounts and ban them. That's why PRC labs that distill from Claude need thousands.

The workarounds used to bypass Anthropic's security measures are quite illegal. They use stolen credit cards, API keys, and accounts. That is only possible in China because any other US/EU lab doing the same would get into massive legal trouble.

That's the moat. Mistral has the capability but not the legal protections.

The 3 US models have their own style.

Qwen3.8 is very clearly distilled from Claude models.

We all get Segways, problem solved.

Add in a heaping dash of anti-american sentiment, and you will get the truth behind the pessimistic commentary.

Downplaying the latest models capabilities is frankly insane considering what we’ve seen what OpenAI’s models have done without safeguards. That wasn’t possible before this latest generation.

> Qwen 3.8 and GLM 5.3 Flash were not just cheaper, the results were significantly higher quality

How can they be higher quality than models they were distilled from?

Unusual means? It’s illegal to steal accounts and credit cards. Completely sanctioned by PRC. No other country can compete since they must follow the law, hence mistral not even trying anymore.

> these proxy services create thousands of new accounts using false identities, fake or stolen credit cards, and stolen API keys. They will often use stolen API credentials belonging to legitimate companies or individuals to give unauthorized entities access to US frontier models.

> Framing learning from observation as somehow bad.

How about hacking accounts, using stolen credit cards, and sending data to the US silently?

> Pretending that this is bad

It is. This makes AI a 2 horse race because European labs cannot legally compete. I find it strange that people from EU cheer for that, it prevents you from having frontier level AI sovereignty.

Except they use thousands of stolen credentials which is definitely not legal. They also don’t notify users they send data to US labs which can include sensitive information.

EU labs like mistral cannot legally do any of these things. So people cheerleading China for it is strange.

They mean distillation is legitimate when labs use one of their own stronger models to train a smaller one. They certainly aren’t advocating for PRC labs to distill Claude for open weight models.

The list of professions made obsolete is very long. Why should programming be protected if AI in the future can make software better, safer, and cheaper?

Yes, I have Tailscale on my AppleTV. Only traffic attempting to reach your tailnet is routed to you, the rest is on his own network.

If you want to make it super easy for non technical people also use reverse proxies to your tailnet so jellyfin/plex is reachable at like “https://watch.example.com”. Pretty much the only friction for non technical people would be creating a tailscale account and being invited to tailnet I guess. But you could also just do that for them I suppose with auth keys

If they could harness the energy from all your atoms it would be worth significantly more than 2 million.

I hate it because I hate having to look at these AI generated images. In fact I want an internet where I have to explicitly opt-in to see AI generated content in general.

My apologies then. Those debrid services seem like they are run in on borrowed time though so I don’t use it. I have friends that do and seem to like it, but I prefer having a local library (*arr stack)

Tailscale. Your friend just needs to be invited to your tailnet. Super easy to use with clients on all the major devices.

This seems like spam in order to get people to sign up to a service with your referral link. It’s literally just a guide on how to create an account.