Kimi K2-Instruct-0905 is an advanced mixture-of-experts language model featuring 1 trillion parameters, with 32 billion activated, designed to enhance coding intelligence and frontend programming experiences. It boasts a doubled context length of 256k tokens, significant performance improvements on various benchmarks, and strong tool-calling capabilities for effective user interaction.

Nvidia has released the Nemotron-Nano-9B-V2, a small language model with 9 billion parameters, optimized for deployment on a single Nvidia A10 GPU. It features a unique toggle for AI reasoning, allowing users to manage internal reasoning and improve performance across various languages and applications.