Mistral 3 - A family of frontier open-source multimodal models
by•
Mistral 3 includes three state-of-the-art small, dense models (14B, 8B, and 3B) and Mistral Large 3 – our most capable model to date – a sparse mixture-of-experts trained with 41B active and 675B total parameters. All models are released under the Apache 2.0 license.
The Ministral models represent the best performance-to-cost ratio in their category. At the same time, Mistral Large 3 joins the ranks of frontier instruction-fine-tuned open-source models.
Replies
DiffSense
Im really impressed with GPT OSS 120B in Antigravity. Its suppose to be level with gpt5 etc. And its possible to run on mac studio, with the highest specs etc. So future look good for local Agentic LLM's. would love to compare this to the GPT OSS model. A local cursor IDE would be huge. Its kind of sad that coding is pay to play at the moment. I spend 100$ a day in coding. It's wild.
Camocopy
I personally tried Mistral 3 Large when it came out yesterday and think it's a good model in terms of multilanguage and multimodality (as even the smaller models with less params can run purely in browser and confidently detect objects correctly - so kudos for that). But the Mistral 3 Large itself still seems to need a bit more power as in my own evaluations it still lags a bit behind the new deepseek 3.2 in logical thinking, coding and math tasks. But yeah, for users in european countries needing a multilingual assistant it's still the best in that scenario
Triforce Todos
Which industries or applications do you think will benefit most from these open models?
Which industries or types of work do you think will get the most help from these open models?