Kimi K3
- Context
- 1.05M
- 1.05M max output
- Input price
- $3.00
- per 1M tokens · Moonshot AI Platform
- Output price
- $15.00
- per 1M tokens
- Epoch Capabilities Index
- 157.7
- #12 of 71 listed
- LMArena Text
- 1,488
- #12 of 79 listed
Where to run it
2 offerings · prices as listed| Meter | Tier | Price |
|---|---|---|
| Input | Standard | $3.00 / 1M tokens |
| Output | Standard | $15.00 / 1M tokens |
| Cache read | Standard | $0.30 / 1M tokens |
| Cache write | 5-minute cache | $3.00 / 1M tokens |
| 1-hour cache | $6.00 / 1M tokens | |
| Input | Standard | ¥20.00 / 1M tokens |
| Output | Standard | ¥100.00 / 1M tokens |
| Cache read | Standard | ¥2.00 / 1M tokens |
| Cache write | 5-minute cache | ¥20.00 / 1M tokens |
| 1-hour cache | ¥40.00 / 1M tokens | |
Changes
All changes →- 2 new prices on Moonshot AI Platform Cache write (5-minute cache) $3.00 · Cache write (1-hour cache) $6.00
Vendor-reported scores
Published by Moonshot AI, under the conditions it states| Benchmark | Setting | Score |
|---|---|---|
| GPQA Diamond | reasoning effort max | 93.5% |
| Humanity's Last Exam | full set, no tools, reasoning effort max | 43.5% |
| full set, with general tools, reasoning effort max | 56% | |
| Terminal-Bench 2.1 | Kimi Code harness, reasoning effort max | 88.3% |
| BrowseComp | context compaction at 300K tokens, reasoning effort max | 91.2% |
| full 1M-token context, no context management, reasoning effort max | 90.4% | |
| MCP-Atlas | 500-task public subset, 100-turn limit, Gemini 3.1 Pro judge, reasoning effort max | 84.2% |
| OSWorld-Verified | reasoning effort max | 84.8% |
| MMMU-Pro | no tools, avg of 3 runs, reasoning effort max | 81.6% |
| with Python tool, avg of 3 runs, reasoning effort max | 83.4% |
Cited scores
From third-party leaderboards, not measured by us- Epoch Capabilities Index 157.7
#12 of 71 listed models · CI 155.1–160.7
Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Sep 26, 2026 · changes we made
- LMArena Text 1,488
#12 of 79 listed models · max effort · CI 1,483–1,493 · 26,400 votes
LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Sep 25, 2026 · changes we made
- Epoch AI · FrontierMath Tiers 1–3 72.2%
#16 of 56 listed models · max effort · ±2.66
Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Sep 26, 2026 · changes we made
- Epoch AI · GPQA Diamond 93.1%
#13 of 64 listed models · max effort · ±1.49
Also listed: high effort 91.9% · low effort 84.8%
Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Sep 26, 2026 · changes we made
- LMArena Text · Chinese 1,540
#7 of 79 listed models · max effort · CI 1,526–1,554 · 1,940 votes
LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Sep 25, 2026 · changes we made
- LMArena Text · Coding 1,541
#6 of 79 listed models · max effort · CI 1,533–1,548 · 6,878 votes
LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Sep 25, 2026 · changes we made