Moonshot AI launches Kimi K3 to challenge OpenAI and Anthropic
Technology
Moonshot AI, a Chinese startup, just dropped Kimi K3, its biggest and boldest AI model so far. It's designed to rival OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5.
While Moonshot admits it hasn't fully caught up yet, it's proud of faster response times and better hardware efficiency.
Kimi K3 tops SWE Marathon
Kimi K3 topped the SWE Marathon test with a score of 42.0, beating both GPT-5.6 Sol (39.0) and Claude Fable 5 (35.0).
But for reasoning skills, Claude Fable 5 still leads the pack, and GPT-5.6 Sol has a slight edge in coding tasks.
The real flex? Kimi packs nearly three trillion parameters, the largest open-weight model so far, and can handle massive prompts thanks to its 1 million-token context window.