Kimi K2

by Moonshot AI

+
+
+
+
About

Kimi K2, released by Moonshot AI on July 11, 2025, is an open-weight Mixture-of-Experts large language model with 1 trillion total parameters and 32 billion active parameters per inference. It features a 256K token context window (expanded from 128K in an September 2025 update) and demonstrated strong performance on agentic coding benchmarks. Kimi K2 targets software engineering agents, tool-use workflows, and open-source deployments under a modified MIT license.

+
+
+
+
Pricing Range
Input (per 1M)$0.60 -$0.60
Output (per 1M)$2.50 -$2.50
Providers1
+
+
+
+
Timeline
ReleasedJul 11, 2025
+
+
+
+
License & Family
License
Modified MIT
Performance Overview
Performance metrics and category breakdown

Overall Performance

2 benchmarks
Average Score
44.1%
Best Score
64.3%
High Performers (80%+)
0

Performance Metrics

Max Context Window
264.2K

Top Categories

Agents
64.3%
Tool Use
23.9%
+
+
+
+
All Benchmark Results for Kimi K2
Complete list of benchmark scores with detailed information
τ-bench
Agents
64.30
64.3%
Unverified
MCP-Atlas
Tool Use
23.90
23.9%
Unverified