Mistral is BACK! (Le Chonk)
Mistral has launched Mistral Large 4, an unofficially named Lechonk model featuring one trillion total parameters with 49 billion active parameters. This hybrid mixture-of-experts architecture delivers multimodal capabilities and competitive cybersecurity performance on the Cybergym benchmark. Pricing is set at $1.36 per million input tokens and $4.18 per million output tokens for the public preview. While the model outperforms other US and European open weights, it remains behind Chinese open source leaders and US closed source frontier models like GPT 6.1 and Opus 5.5.
Top 3 new model launches at Gemini Audio at Night
Google DeepMind has unveiled three groundbreaking audio models designed to transform live translation, voice personalization, and developer integration. The new text-to-speech system enables precise emotional control and resonance customization while supporting real-time translation across over one hundred languages. Gemini Live models now incorporate multimodal understanding and function calls directly into developer projects for seamless interaction. These tools allow users to generate specific voices and process audio or video feeds with unprecedented flexibility and accuracy.