Kimi
Moonshot AI · China · Language & multimodal
China Open weights
Showing version: Kimi K2 Thinking · 2025-11-06 · Version lineage ↓
01
Overview
Kimi is Moonshot AI's model family; K2 is a 1T-total / 32B-active MoE open-weight model focused on coding and agentic tasks, with K2 Thinking adding interleaved reasoning and tool calls.
Context length262,144 tokensInput$0.60 USD / 1M tokensRelease date2025-11-06
02 / Provenance
Field checks and source notes
When sources disagree on a field, every reading is listed here with its own citation. We do not take a midpoint, invent a missing figure, or decide which side is right — follow the citation and judge for your case.
These field-level notes are recorded verbatim in Chinese in the source data. Open the Chinese version
03 / Plain layer
Plain-language benchmarks
Reasoning 44.9%Humanity's Last Exam, text only, with tools; heavy mode 51.0 and no-tools 23.9 listed separately on same source, not merged
Coding 71.3%with tools, INT4 precision per Moonshot model card reporting; maker-reported, not the swebench.com independent run (that run lists 63.4% for K2 Thinking on same page, not merged)
Memory 262,144 tokens
Model size Not disclosed
Price / free $0.60 / $2.50 USD / 1M tokens
04
Practical facts
| Context length | 262,144 tokens |
|---|---|
| Max output | Not disclosed |
| Parameters | Not disclosed |
| Input | $0.60 USD / 1M tokens |
| Output price | $2.50 USD / 1M tokens |
| CNY | Not disclosed |
| Free tier | Not disclosed |
| Output speed | Not disclosed |
| Release date | |
| License | Not disclosed |
Show technical data
Technical benchmark table
| Benchmark | Score | Max / unit | Source type | Source | Test date | Conditions |
|---|---|---|---|---|---|---|
| SWE-bench Verified swe-bench-verified | 71.3% | percent / 100 | Vendor-reported | with tools, INT4 precision per Moonshot model card reporting; maker-reported, not the swebench.com independent run (that run lists 63.4% for K2 Thinking on same page, not merged) | ||
| HLE hle | 44.9% | percent / 100 | Vendor-reported | Humanity's Last Exam, text only, with tools; heavy mode 51.0 and no-tools 23.9 listed separately on same source, not merged |
| Version | Release date | Status | Openness | Parameters | What changed | Detail | |
|---|---|---|---|---|---|---|---|
| Kimi K2-0905 | Active | Open weights | Not disclosed | TBD | Detail | ||
| Kimi K2 | Active | Open weights | Not disclosed | TBD | Detail | ||
| Kimi K2 ThinkingShowing version | Discontinued | Open weights | Not disclosed | TBD | Detail |
Last updated: · Sources on detail page
06 / Compare