Moonshot AI released Kimi K3, a 2.8-trillion-parameter model with built-in image understanding and a one-million-token context window, available today in the Kimi apps, the Kimi Code terminal tool and the Kimi API, with full model weights to follow. The company says it leads other tested open models on its evaluation suite but still trails Claude Fable 5 and GPT 5.6 Sol overall. It also warns that the model can act on its own initiative when instructions are ambiguous, and that quality degrades if a session is switched to it mid-way.
What changed
The largest openly released models were smaller, and Kimi's previous flagship was K2.
What it unlocks
Running or calling a near-frontier model with a one-million-token context and image understanding, with weights published for self-hosting.
- 2.8T parameters, 16 of 896 experts active
- 1M-token context window
- $3.00/MTok input, $15.00/MTok output
- $0.30/MTok cache-hit input
What you need to act on it
- for self-hosting, 64 or more accelerators recommended
- API account for hosted use
- weights due by 27 July 2026
- kimi.com2026-07-31