LLM Releases
← Catalog

Kimi K2 Thinking

Available
Moonshot AIFrontierOpen weights

Open K2 reasoning-agent variant that interleaves step-by-step thinking with tool calls and supports stable 200-300 step tool-use trajectories.

Specifications

License
Open weights · Modified MIT
Weights
Downloadable
Architecture
Mixture-of-Experts
Parameters
1T · 32B active
Context window
256K tokens
Max output
32K tokens
Knowledge cutoff
Price (in / out, $/M)
Modalities
TextCode

Benchmarks

GPQA Diamondself-reported
84.5%
source ↗
SWE-bench Verifiedself-reported
71.3%
source ↗
AIME 2025self-reported
94.5%
source ↗

Vendor-reported figures are claims until independently verified. See methodology.