Kimi K3 is Moonshot AI's July 2026 flagship model, a 2.8-trillion-parameter mixture-of-experts model with native multimodal vision and a 1M-token context window. It is built for long-horizon coding, reasoning, and agent workflows, positioned as an open-weight competitor to top proprietary frontier models.
Cost rate
2.5x
Context
1M
Released
Jul 16, 2026
Input
TextImageVideo
Output
Text
Support
ReasoningTool callingStructured outputs
Performance
Median latency and throughput measured across recent requests.
Best at
The categories where this model ranks highest.
Legal & Government
#12 路 top 3%
Life, Physical, & Social Science
#13 路 top 3%
Medicine & Healthcare
#13 路 top 3%
Mathematical
#17 路 top 4%
| Domain | Score | Rank |
|---|---|---|
| CodingCode Arena Elo | 1658 | #23 |
| AgenticAgent Arena score | -0.00 | #37 |
| Software & IT Services | 1523 | #20 |
| Writing, Literature, & Language | 1475 | #27 |
| Life, Physical, & Social Science | 1517 | #13 |
| Entertainment, Sports, & Media | 1461 | #28 |
| Business, Management, & Finance | 1489 | #18 |
| Mathematical | 1511 | #17 |
| Legal & Government | 1506 | #12 |
| Medicine & Healthcare | 1512 | #13 |
| Category | Win rate | Rank |
|---|---|---|
| ASCII Art | 68.3% | #3 |
| Data Visualization | 63.6% | #4 |
| Website | 59.5% | #5 |
| Code Categories | 60.4% | #6 |
| SVG | 59.6% | #8 |
| Game Development | 55.5% | #10 |
| UI Component | 60.7% | #11 |
| 3D | 63.6% | #13 |
Latest news about Kimi K3
Recent articles and community posts, linked to their original sources.
- Kimi-K3 Technical Report [pdf]Hacker News391183
Kimi-K3 on HuggingFaceHacker News1,382544
Moonshot AI suspends new subscriptions due to Kimi K3 demandHacker News284114
Testei o Kimi K3.1 sem fazer joguinhomano deyvin77,016- China鈥檚 moonshot AI puts Kimi models into race for scientific discoveriesSouth China Morning Post on MSN
Kimi K3.1 Pro model weights are outHacker News641191
Kimi K3.1 Pro explained in 6min..Caleb Writes Code56,283
Moonshot Kimi K3.1 Pro is Trash? My Honest ReviewSara on AI1,630