Mistral 3 family of models released
mistral.ai
Mistral 3 family of models released
1–10 of 243 posts
Re: Mistral 3 family of models released
#2Extremely cool! I just wish they would also include comparisons to SOTA models from OpenAI, Google, and Anthropic in the press release, so it's easier to know how it fares in the grand scheme of things.
Re: Mistral 3 family of models released
#3Do all of these models, regardless of parameters, support tool use and structured output?
Re: Mistral 3 family of models released
#4I still don't understand what the incentive is for releasing genuinely good model weights. What makes sense however is OpenAI releasing a somewhat generic model like gpt-oss that games the benchmarks just for PR. Or some Chinese companies doing the same to cut the ground from under the feet of American big tech. Are we really hopeful we'll still get decent open weights models in the future?
Re: Mistral 3 family of models released
#5Upvoting for Europe's best efforts.
Re: Mistral 3 family of models released
#6Extremely cool! I just wish they would also include comparisons to SOTA models from OpenAI, Google, and Anthropic in the press release, so it's easier to know how it fares in the grand scheme of things.
The lack of the comparison (which absolutely was done), tells you exactly what you need to know.
Re: Mistral 3 family of models released
#7Looks like their own HF link is broken or the collection hasn't been made public yet. The 14B instruct model is here:
https://huggingface.co/mistralai/Ministral-3-14B-Instruct-25...
The unsloth quants are here:
https://huggingface.co/unsloth/Ministral-3-14B-Instruct-2512...
Re: Mistral 3 family of models released
#8This is big. The first really big open weights model that understands images.
Re: Mistral 3 family of models released
#9Do all of these models, regardless of parameters, support tool use and structured output?
In principle any model can do these. Tool use is just detecting something like "I should run a db query for pattern X" and structured output is even easier, just reject output tokens that don't match the grammar. The only question is how well they're trained, and how well your inference environment takes advantage.
Re: Mistral 3 family of models released
#10Extremely cool! I just wish they would also include comparisons to SOTA models from OpenAI, Google, and Anthropic in the press release, so it's easier to know how it fares in the grand scheme of things.
The lack of the comparison (which absolutely was done), tells you exactly what you need to know.
If someone is using these models, they probably can't or won't use the existing SOTA models, so not sure how useful those comparisons actually are. "Here is a benchmark that makes us look bad from a model you can't use on a task you won't be undertaking" isn't actually helpful (and definitely not in a press release).