Mistral
French AI company offering open-weight language models, the Le Chat assistant, and an API.
Liked for cheap small models that suit OCR and data work,12 strong Voxtral speech transcription,34 and EU-based deployment.56 The most common complaint is that its models trail competitors on quality and value,789 and several commenters object that its license is not truly open.101112
Tiers 91 people recommend Mistral across these; 41 of them said just "Mistral"
By version sentiment of people who named that release, latest release first
Mistral Large 2 versions, latest 4
Versions with at least 5 people. Small samples swing a lot, so read the counts too. Compare all models →
Why people recommend it
Low cost · 3
Commenters report paying very little for OCR, data cleaning and other API workloads.
“I do a lot of OCR, file analysis, stuff like that. I use Mistral for that.”
“I used Mistral 3. It costed me literally pennies.”
Show 1 more
Local running · 3
Smaller models and Devstral run usefully on personal hardware, including an 8GB GPU laptop.
“fast enough that I never wait for it”
“I run Ministral-3b on my hardware as my go-to”
Show 1 more
“A competitive coding like devstral 2 runs fast enough to be very helpful”
Speech models · 3
Voxtral is praised for transcription accuracy, including acronyms and technical jargon in meeting transcripts.
“No other model, open or closed, has been able to hit such a low AER (Acronym Error Rate ;)) for my meeting transcripts.”
“Their Voxtral[1] speech models are really good.”
Show 1 more
“In my tests with other languages Voxtral was better then Whisper.”
Everyday use · 3
Some commenters use it as their main assistant for coding, research and general questions.
“I use it as my workhorse for coding and general chat questions, because it's good enough 80% of the time”
“Mistral performs well for me and I see no reason to use something else.”
Show 1 more
“I am absolutely amazed how good it is.”
European option · 2
EU-based commenters value it as a European provider that can also be deployed on-premise.
“Mistral has been great for us as a provider for our EU-based company. We use Mistral Large 4, GLM 5.3 Flash and DeepSeek v4.1 Flash 1017 through their API”
“they are a viable solution for your typical European enterprise”
Data tasks · 3
Commenters report good results on OCR, file analysis, formatting and data cleaning.
“I do a lot of OCR, file analysis, stuff like that. I use Mistral for that.”
“the fact it's a "Medium" and 5x the price of the Large definitely concerns me”
Show 1 more
Caveats and dissent
Competitiveness · 3
Critics say its models fall behind rivals on capability and on intelligence for the cost.
“I don't think any Mistral model can match an open 30B-sized Qwen from a year ago, so right now they're not really a competitive alternative to anything at all.”
“Mistral 4 is not competitive.”
Show 1 more
“Mistral was amazing at first, Le Chat. Almost as good, generous free limits, good docs. Now? Just plain bad.”
Licensing limits · 3
Commenters dispute calling Mistral Large 4 open, and some regret that the newer Voxtral has no open weights.
“Calling this "open weight" is misleading.”
“Very much not open as in open source.”
Show 1 more
Data privacy · 2
An organization buyer reports prompt training defaulting to on, even on the Team tier, with central opt-out seemingly lost.
“after being disappointed that the Pro tier was opt-in to training on prompts by default”
“Mistral now trains on user input by default, except on enterprise tier”
Rising prices · 3
Commenters flag steep price increases on newer models, and one still finds Large 4 too expensive.
“priced $1.5/$7.5 per million input/output tokens (change from the previous $0.4/$2)”
“the fact it's a "Medium" and 5x the price of the Large definitely concerns me”
Show 1 more
“it's still too expensive and very slow when thinking.”
Reliability · 2
One commenter calls Mistral models dangerous unless closely reviewed, and another found Mistral Small 4 too buggy to use.
“My biggest issue with Devstral and even their biggest model is that they’re dangerous unless closely directed and reviewed”
“MistralSmall4 is too buggy to use”
Coding tools · 2
Commenters were disappointed by the Mistral Vibe harness and by Codestral for autocompletion.
“I feel sorry for whatever they did to that model”
“We gave it a shot and it was lackluster at best..”
Often compared with
1 moved from Claude to Mistral
1 moved from Gemini to Mistral
1 moved from Mistral to Gemma
Who recommends it in their own words
Threads where it came up
- Mistral 3 family of models released 7 mentions
- Mistral raises €3B 5 mentions
- Leanstral 1.5 4 mentions
- Notes from the Mistral AI Now Summit in Paris 4 mentions
- Can I opt out of my input or output data being used for training? 3 mentions
- Grok 4.3 3 mentions
Also written as Mistral 7B, Mistral 3, Mistral AI, Mistral 3 3B. Data: mistral.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 49
“I am absolutely amazed how good it is.”
“Mistral has been selling GLM5.3 inference hosted in Europe on their own hardware”
“Mistral has a free moderation endpoint with ZDR. Can definitely recommend it!”
“Essentially you are right, I could stack subscriptions via a Team plan, but for a premium.”
“Definitely the best I've seen from any Mistral model”
“Would love to use Mistral more (as EU alternative), but my last test (though quite a while ago) I was a bit disappointed.”
“powered by Mistral in the EU”
“Mostly I use it with Mistral”
“I don't think any Mistral model can match an open 30B-sized Qwen from a year ago, so right now they're not really a competitive alternative to anything at all.”
“Read this https://europe.mistral.ai/ if you can bare to go through the slop, has it's share of regulatory capture strategy.”
“they are a viable solution for your typical European enterprise”
“I love Mistral and I use it whenever I don't care about quality.”
“have a good EU alternative to Opus or even Sonnet”
“I used Mistral 3. It costed me literally pennies.”
> being disappointed that the Pro tier was opt-in to training on prompts by default "Opt-in" means that the default is non-participation, for example, not training on your prompts. Is it possible that you intended to say "opt-out", which means that the default is participation? That's what the context seems to suggest. See, for example, https://termly.io/resources/articles/opt-in-vs-opt-out/: > Data privacy laws like
“Mistral now trains on user input by default, except on enterprise tier”
“after being disappointed that the Pro tier was opt-in to training on prompts by default”
This is not a patent application. This is granted. They applied in March and was fast-tracked until June. This is a full grant. You can see the file claims here: https://ibb.co/k2mVVkFN
“It integrates with Mistral”
“Another cool thing about Mistral 7b is it runs offline”
“I once tried to index the web with local Mistral 7b.”
“Mistral: eq.IQ ~70”
“So I signed up with Mistral instead.”
“With some outliers like Mistral and Llama on some questions.”
“Mistral's multiple products sections actually look good, lot can be done there”
“Now I use just Mistral and local Vibethinker-3B, but they work just fine.”
“I do a lot of OCR, file analysis, stuff like that. I use Mistral for that.”
“I used them because they had the fastest chat response.”
“I use it because EU and API pricing is decent to me. And support is awesome also.”
“Mistral is a bonkers weird writer.”
“I use it as my workhorse for coding and general chat questions, because it's good enough 80% of the time”
“Mistral writes so weird it almost trips me up”
“Europe has Mistral.”
“maaaybe sometimes Mistral for a simple EU subscription”
“it is so far behind the American models that I just never bother using it”
“Mistral, ymmv”
“EU NatSec uses cases tending to use Mistral”
“so never tested Mistral on that front”
“Mistral doesn't count because it seems to have refused to participate”
“mistral was both faster, cheaper, and substantially better than any alternative”
“Mistral Large 4 is definitely not open”
“Calling this "open weight" is misleading.”
“it's still too expensive and very slow when thinking.”
“Mistral has been great for us as a provider for our EU-based company. We use Mistral Large 4, GLM 5.3 Flash and DeepSeek v4.1 Flash 1017 through their API”
“Mistral 4 is not competitive.”
“Very much not open as in open source.”
“No other model, open or closed, has been able to hit such a low AER (Acronym Error Rate ;)) for my meeting transcripts.”
“Their Voxtral[1] speech models are really good.”
“it performed the best in my benchmarks”