YHK - yet another hacker news clone

Mistral Large 4

231 points funnym0nk3y 1 hour ago 57 comments

moecables 1 hour ago | parent

compared to Deepseek 1017, I'm not sure why I'd use this?

https://openrouter.ai/compare/mistralai/mistral-large-4-0/de...

serioussecurity 1 hour ago | parent

If you're required to use a European model

walrus01 1 hour ago | parent

Interestingly Mistral has been selling GLM5.3 inference hosted in Europe on their own hardware, for those who are OK using a "not developed in Europe" model but run on hardware located in and controlled by a European corporation.

https://docs.mistral.ai/models/zai-glm-5-3

frotaur 1 hour ago | parent

Indeed, I see no difference between a chinese model hosted in Europe vs a Europe-grown model.

You could argue that potentially you can train a backdoor into a foreign-grown model (that has been demonstrated to be possible in toy examples), but I would highly doubt that this would be the case today, given the potential reputational damage if discovered.

debugnik 54 minutes ago | parent

Didn't we get a submission today about Qwen3-Coder producing more bugs when hinted on the user having different ideology than the Chinese government?

Ok I see it's flagged now, probably fake, but the fact that it's possible does make me feel uneasy about both Chinese and US models.

walrus01 50 minutes ago | parent

I recall seeing something about that. I don't know of anyone still using such an ancient model when much newer Qwen 3.5-120, 3.6-35B-A3B, 3.8-27B and 3.8-flash-next all exist and score much higher on all conceivable metrics. I would like to see if anyone can replicate the same thing on a 'current' Qwen released in the past 6 months.

verdverm 48 minutes ago | parent

It was so insignificant, it could be equally described by the non-determinism inherent in the models. And that it reads more like a marketing piece.

That's my read anyhow. It's been flagged. The whole thing is kind of rich, considering the Israeli roots of the company.

https://news.ycombinator.com/item?id=50179706

IshKebab 48 minutes ago | parent

> I see no difference between a chinese model hosted in Europe vs a Europe-grown model.

Really? Fascinating. Do you know about Tiananmen Square, the great firewall of China, etc?

Muromec 25 minutes ago | parent

DS4 hosted by deepseek used through the API doesn't have a problem talking about tiananmen square and Uyghurs, but it thinks it's Claude when it speaks about it. I'm not sure what it all means. That was on v4-pro before the 4.1 switch, so may or may not be reproducable.

Add: I was surprised by the fact that the LLM itself doesn't know it's identity unless the system prompt mentions it.

traceroute66 18 minutes ago | parent

> Do you know about Tiananmen Square

People here bringing up the straw clutching argument of "but Tianament Square" is becoming really tiresome.

The fact is a Chinese LLM's lack of knowledge about Chinese politics will not affect 99.99999999999% of users and their workflows.

And for the 0.000000001% who are affected, they don't need an LLM. A quick search of Wikipedia or 5 seconds on Google will suffice.

What the hell has Tiannamen Square got to do with Joe Blogg's vibecoding session ? Absolutely nothing.

kolinko 44 minutes ago | parent

Remember Stuxnet? An american worm that was completely benign, travelled on usb drives in hiding, and only activated when it guessed that it's on Iranian plants, and iirc changed very subtly some very specific parameters of operations.

Here, with a chinese model, it would be even simpler - you finetune a model to activate certain operations when it discovers certain environments, but to behave completely normal otherwise. Kind of like a sleeper agent. And with current tech we wouldn't even be able to spot this.

traceroute66 22 minutes ago | parent

> And with current tech we wouldn't even be able to spot this.

Meanwhile Anthrophic, OpenAI & Co can't even manage to stop their LLMs from hacking into third-party servers until they discover it after the event.

What happened to people's ability to implement basic security architectures like sandboxing etc ?

Muromec 16 minutes ago | parent

It's not like they didn't sandbox it or limit the access. There were reasonable measures, but not really serious. They made a kodoku pit and didn't watch it.

The agents started hacking because they were sandboxed.

traceroute66 4 minutes ago | parent

> The agents started hacking because they were sandboxed.

Not sure your definition of sandbox but mine doesn't include unfiltered internet access ?

traceroute66 24 minutes ago | parent

> you could argue that potentially you can train a backdoor into a foreign-grown model

If we're wearing tinfoil hats then do we know the NSA have not planted stuff in Anthropic and OpenAI anyway ? ;)

I think the Chinese would be happy enough to see OpenAI and Anthropic suffer because people discover open-weights can get you most of the way there for a fraction of the cost.

You don't have to target the end users. Economic impact on the US providers would be sufficient.

The Chinese are also no doubt seeking to play the long game of hoping to (eventually) rebalance the US-centric world order in an Easterly direction.

traceroute66 33 minutes ago | parent

> for those who are OK using a "not developed in Europe" model but run on hardware located in and controlled by a European corporation.

For the record this is also possible in Europe on Privatemode[1] which also gives you enclave-based processing. And access to GLM-5.3 Flash as well as GLM-5.3

[1]https://www.privatemode.ai/

ex1fm3ta 1 hour ago | parent

if you care about data-retention policy or you have to use european models ...

Systemerror7A69 1 hour ago | parent

Well, Models are hard to just distill down to a single benchmark number, or even a few across multiple different benchmark - they all behave and work slightly differently and you can find some who work or feel better for you and your use-case specifically.

But you're right, Mistral Large 4 seems a much more expensive but overall weaker model across the board.

The reason to use them would be, well, because they're...Mistral. They're in EU, france is large in nuclear energy so the energy cost of training has much less of an environmental impact, the data protection laws are very good in EU. This can really be a deciding factor for many EU companies, and I think it's good that they now have something which is at least semi-usable. (apparently)

It's also good that other options exist, even if they are worse. Deepseek is kind of like a large specter looming over every model release, because it's just SO cheap and good. That doesn't mean we should abandon every other model.

Still, the model is not really competitive on price and abilities alone. I'm mostly happy to see Mistral still in the fight.

petercooper 57 minutes ago | parent

I say this with a slight tongue in cheek, but Mistral does have a better option: they recently added GLM 5.3 to their platform (and it works rather well). Not trained in the EU, though, of course ;-)

Systemerror7A69 35 minutes ago | parent

Yeah, absolutely true. And I am guilty of simply using Mistral as a GLM subscription myself. I just really like the GLM 5.3 model.

I will see where Chonk is better though, since you have both options in one subscription.

andix 58 minutes ago | parent

Patriotism.

lwansbrough 45 minutes ago | parent

Well when your greatest ally acts like an adversary, who can blame them for the increased sovereign focus?

maxloh 16 minutes ago | parent

Maybe not now, but in a few years, we might not need frontier models anymore. Open-weight models are only half a year behind the frontier, and both categories are advancing really fast.

Gemini 2.0 Flash, released just 1.5 years ago, could not even repeat a few paragraphs unchanged. Now, we have Mistral, DeepSeek, and MiMo.

cyanydeez 4 minutes ago | parent

The TAM is already covered by local models. 10% of a health society could use better models.

We are in the 70s era of gas guzzling before the gas embargo and cheap foreign imports and economy cars reign.

yorwba 1 hour ago | parent

Previous discussion: https://news.ycombinator.com/item?id=49977979 (2044 points 15 days ago, 1254 comments)

denysvitali 10 minutes ago | parent

I was confused and thought it's a new model... but no, it's the same. What's the purpose of reposting this as a new story? Am I missing something?

Flux159 1 hour ago | parent

I think that this link is more helpful: https://mistral.ai/news/mistral-large-4/

Seems like it's multimodal and visual intelligence capabilities are better than Deepseek 4.1 Flash, even though coding is not equal to DS 4.1 1017. The cost is about 10x of Deepseek though comparing on OpenRouter.

Weights will be here https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B at the end of the month.

Systemerror7A69 57 minutes ago | parent

It's gonna be interesting to see the final version, since this is a preview model still. I'm not completely sure as to how this works, but from what I gathered it will be further trained during this period, so it might improve a bit.

Since cost is so much higher than deepseek though it really seems like the biggest reason to use it would be "in the mistral subscription" for personal use or "because you need to abide by european data privacy laws" for companies.

Mostly I'm happy they're still in competition, as I mentioned already elsewhere.

erkt 33 minutes ago | parent

Are they still in competition? Or are they a government sanctioned monopoly? I’m not sure that will be competitive. There are no market forces to do so but pride.

OKRainbowKid 22 minutes ago | parent

Which government? Sanctioned how? What stops other European companies from competing with Mistral? Is the situation better in the US?

jxf 18 minutes ago | parent

A lot of national defense stuff is a de facto "government sanctioned monopoly" and this is certainly the same flavor. Market forces are different if the industry is perceived to be vital to continued sovereignty.

0gs 16 minutes ago | parent

french are pretty proud man. pro-euro europeans too!

olivierduval 14 minutes ago | parent

Actually, with Trump in the Office, "AI/Data Sovereignty" is quite a strong market forces. From Europe point-of-view, any US company is a potential risk (hello Starlink)

;-)

hnfong 9 minutes ago | parent

Being the only(?) competitive European model probably makes them a de-facto monopoly for doing sensitive European stuff.

maelito 7 minutes ago | parent

It's well known that every Mistral employee is a fonctionnaire. No.

andix 59 minutes ago | parent

I see Le Chaton Fat and press upvote, sure. But why? Wasn't it released already 2 weeks ago?

wojtek1942 49 minutes ago | parent

Some details and benchmarks were announced but the weights didn’t get released

ismailmaj 40 minutes ago | parent

Still 9 days before the release of weights apparently.

Betelbuddy 47 minutes ago | parent

Le Chaton Fat is a lovely name. I predict the next Michelin restaurant with will have an LLM as the Chef...

nik736 42 minutes ago | parent

The weird thing about this model is that it spits out Chinese characters just like the chinese models when chatting with it in German. Happened to me quite often during the past few days.

bigwheels 34 minutes ago | parent

Does this suggest Mistral is distilling Deepseek et. al.? Can distillation keep passing through with high fidelity, or is it lossy like copying a VHS tape?

orbital-decay 20 minutes ago | parent

It doesn't, even remotely. It's a typical failure mode in all models, some do this more often. It's not related to samplers either. Not necessarily Chinese characters either, they likely just occupy a large portion of vocabularies. For example Gemini 2.0 used to glitch in Telugu and Ukrainian, Claude 2.x and 3.0 used to switch between random languages for a few words when writing in non-English, once in a blue moon.

speedyStuff_ 27 minutes ago | parent

Opus 4.8 and Fable 5 did the same; they would output random Chinese characters and then say “I made a typo, let me fix it” and it’s just them removing that stay Chinese characters. Haven’t seen this since 5.1 family of models.

paulddraper 24 minutes ago | parent

I saw it in earlier Claude models, but I haven't seen it for a least 7-8 months.

c0rruptbytes 22 minutes ago | parent

i’ve seen it and Cyrillic in the 5x family

dadoum 21 minutes ago | parent

Google also does that surprisingly

h1fra 19 minutes ago | parent

you can see the same thing with any models

bmk44 17 minutes ago | parent

Ha, ha. Just like when Tilly Norwood switched to Chinese while interviewing with Piers Morgan [1].

On a serious note, I remember getting spooked when I saw a bunch of Chinese characters dumped out into some of my server logs (pre-LLM era). When I pasted the Chinese content from my log into google translate, it was just complete nonsense. Later I found a bug that was dumping raw binary data into my log file which interpreted as Chinese characters when viewed in a text editor. Perhaps similar bugs existed in the models.

[1] https://www.youtube.com/watch?v=8VOfGcTqqq8

whiplash451 11 minutes ago | parent

Can you please share: - what specific input you see this issue with? - which interface you used (API, playground, specific sdk version or a wrapper app)? - whether the chinese chars appear in the thinking section or the output?

Thanks.

dropbox_miner 42 minutes ago | parent

waiting for pelican bench results (the only benchmark that matters)

dropbox_miner 40 minutes ago | parent

NVM. Looks like it already exists:

https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

simonw 11 minutes ago | parent

Yeah I did this one two weeks ago when it was first released. Not sure why it's on the HN homepage today.

maelito 5 minutes ago | parent

Because people acclaim European competition on the LLM scene, I guess. We don't have much to show.

maelito 41 minutes ago | parent

What's more impressive is the datacenter used for training. It shows you can get "sub-SOTA" for one tenth of the resources of SOTA.

skohan 40 minutes ago | parent

Impressive if these benches hold up to real world use. Not groundbreaking, but certainly in the game.

Would love to see them release some smaller open-weight models.

codeulike 33 minutes ago | parent

"Covered by GDPR" is a very big plus for me. Also very simple training opt-out.

jakswa 29 minutes ago | parent

52B active parameters and my qwen-3.8-flash-next model running at home is supposedly competitive? maybe I gotta test it to believe it. https://artificialanalysis.ai/models/comparisons/mistral-lar...