Mistral AI Introduces Large 4 Model
Mistral AI has launched its latest AI model, Large 4, featuring an impressive 1 trillion parameters, on October 6. This model activates 49 billion parameters per query and is priced at $1.36 per million input tokens and $4.18 per million output tokens, making it one of the more affordable options in the market.
Playful Naming and Origins
The model's whimsical name, "Le Chonk," is inspired by the "Le Chaton Fat" meme that circulated in June. Mistral's CEO, Arthur Mensch, humorously noted that the name translates to "the big kitten" and reflects the model's playful branding. The company plans to release the model's weights by the end of October for public access.
Performance Metrics
In preliminary evaluations, Large 4 scored 54.7 on the Finance Agent v2 benchmark, surpassing GPT-6 Astra's score of 53.5 but not quite reaching Claude Opus 5.5, which scored 58.6. Mistral's model utilizes a "mixture of experts" architecture, allowing it to activate a smaller subset of parameters for each query, enhancing efficiency compared to its predecessor, Large 3.




