I wonder why Mistral with Mistral Small 24B and Google with Gemma 3 27B aren't comparing against Qwen 2.5 32B in their press releases. As far as I understand it is a clear competitor in capability, model size and licensing of the weights.

That said: great to see a multimodal take on Mistral Small and it's really exciting to have more strong options in the 14B-32B range.

Or Phi-4. Which seems to outperform it and be a smaller model from 4 months ago.

Small bump to evals from Mistral Small 3 -> Mistral Small 3.1

https://lifearchitect.ai/models-table/

I love this size for my 64GB MacBook Pro. I’ve been going back and forth between Mistral Small and Gemma 3. I love how quick this model is.