Kimi K3 from Moonshot AI is here. It’s a massive 2.8 trillion-parameter open-weight multimodal reasoning model with a 1M-token context window, built for complex coding, knowledge work, and long-horizon agentic workflows.

Why Kimi K3 matters

• 2.8T parameters — one of the largest open-weight reasoning models available
• 1M-token context — fits entire codebases, books, or long research corpora
• Multimodal — processes text and images in the same conversation
• Agentic — engineered to drive multi-step tool-using workflows without losing the plot

Where it fits on Chat-O

Kimi K3 is available in the Power tier, alongside other frontier reasoning models like DeepSeek R1, Grok 3, and GPT-5. Choose it when you need:

• Deep, multi-file refactors where you want the model to “hold” the whole project
• Long-running agent tasks with many sequential tool calls
• Knowledge-intensive questions grounded in a large uploaded document set

How it compares to Kimi K2.5

Both stay available — K2.5 remains a great Balanced-tier daily driver, while K3 is the choice when you need maximum reasoning depth and context.

Start using it

Select Kimi K3 from the dropdown in any chat. Sign up free to get starter credits, or jump in if you’re already a member — no API keys to manage, no separate subscription.