Kimi K2 Instruct
Non-thinking Kimi K2 MoE instruction model for coding and tool-using agents
- Strongest at
- Writing Top 35% of the past year's models · 2 of 4 benchmarks
- Context window
- 131.1K tokens in · 32.8K out
Capability profile
Scored against 90 models released in the past year. 50 is the average model; further out is better.
No results yet: Reading & documents, Math, Science, Visual reasoning, Computer use
Where these scores come from
Each capability combines the published benchmarks listed here, rescaled so 50 is the reference average and 15 points is one standard deviation. A hollow dot rests on a single benchmark.
- Coding Aider polyglot, GSO-Bench
- Writing EQ-Bench Creative Writing v3, EQ-Bench Longform Writing
- Reasoning SimpleBench
The Kimi K2 Instruct line
2 releases in the Kimi K2 Instruct line.
A newer release is out: Kimi K2 Instruct 0905.
Epoch benchmark boards
Best result on each board it has run. The tick marks the leader.
Moonshot AI's past year
7 release days in the past year, about one every 52 days. The next is expected around Dec 7, 2026. Forecasts ↗
Specifications
Specifications and prices from Models.dev; availability and limits may differ by provider. Benchmarks from Epoch AI (CC BY 4.0), LiveBench and the other sources named above.