Moonshot AI launched the Kimi K3.1-Pro AI model featuring a 1M token context window. The company said it matches top proprietary models on several benchmarks. It handles complex coding and financial workflows with ease. Users can access the model now across various platforms, with full weight releases expected by November 4.
The Kimi K3.1 Pro is an enhanced version of its predecessor, the Kimi K3.
It was built with significantly more computing resources during post-training, according to Moonshot AI.
The company claims that this new model can handle all types of coding work, from quick fixes in real codebases to full-project builds and even end-to-end delivery of complete financial workflows.
Moonshot AI has optimized its reinforcement learning (RL) framework, long-context processing, training efficiency, and infrastructure to boost the model's performance across hundreds of application areas.
The company claims that the Kimi K3.1 Pro matches the performance of top proprietary AI models on several benchmarks.
It is now available across the company's platforms including Kimi.ai, Kimi mobile apps, Kimi Work and Kimi Code, as well as through the Kimi API.
The high-speed version of the model, Kimi K3.1 Pro Highspeed, is 2-3 times faster than its predecessor.
This makes interactions and tasks much snappier for users.
The full access to the model's weights is expected to be released by November 4, giving users even more ways to leverage this advanced technology across Moonshot AI's applications.