Today we are releasing two small models: Mathstral 7B and Codestral Mamba 7B. On the MATH benchmark, Mathstral 7B obtains 56.6% pass@1, outperforming Minerva 540B by more than 20%. Mathstral scores 68.4% on MATH with majority voting@64, and 74.6% using a reward model. Codestral
@guillaumelample
-
Mistral Fine-Tuning API Launches for Model Customization
By
–
Mistral fine-tuning API is out ! You can now fine-tune your own Mistral models and deploy them efficiently on La Plateforme : https://
mistral.ai/news/customiza
tion/
… In many cases, fine-tuning allows small models to match (and sometimes surpass) the performance of much larger models, but with -
Codestral becomes second most popular model on Mistral chat
By
–
good thread: https://
reddit.com/r/LocalLLaMA/c
omments/1d86di0/codestral_solved_a_problem_in_two_messages_that_i/
… Codestral is now the second most popular model on https://
chat.mistral.ai (and it's free!) -
Mistral AI Releases Codestral-22B Model Weights on HuggingFace
By
–
We are also releasing the weights on HuggingFace! https://
huggingface.co/mistralai/Code
stral-22B-v0.1
… More details on our blogpost: -
Codestral-22B: New Code Model Outperforms Larger Models
By
–
Today we are releasing Codestral-22B, our first code model! Codestral is trained on more than 80 programming languages and outperforms the performance of previous code models, including the largest ones. It is available on our API platform, through instruct and
-
Mistral AI Team at ICLR-2024 Vienna Conference
By
–
We will be at ICLR-2024 in Vienna this week with a few people from the @MistralAI team. Feel free to DM if you want to catch up!
-
Mistral Large and Mixtral 8x22B Performance on GSM1K Dataset
By
–
Nice paper. Very good performance by Mistral Large and Mixtral 8x22B on the new GSM1K dataset! Results from Table D (plotted in Figure below) are also a good reminder that models are very sensitive to prompt. In general, it is also good to use sampling & majority voting to
-
Congratulations to Mistral AI team on their amazing release
By
–
Congratulations to all the @MistralAI team for their amazing work on this release!
n/n -
Mixtral 8x22B Instruct Model Released on Mistral Platform
By
–
Mixtral 8x22B Instruct is available on our platform, along with all our models: https://
console.mistral.ai The base and the instruct models are also both available for download on Hugging Face ! https://
huggingface.co/mistralai/Mixt
ral-8x22B-v0.1
… https://
huggingface.co/mistralai/Mixt
ral-8x22B-Instruct-v0.1
… 3/n -
Mixtral excels in science, code, and multilingual tasks
By
–
Mixtral particularly shines in science, code, and multilingual (below are performance results for the base model). 2/n