Kimi K3
k3
Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work
Specification
most-agreed values
Context
1M
Max output
131K
Released
2026-07-16
Knowledge cutoff
—
Retires
—
Open weights
yes
Input
text, image, video
Output
text
Price
US dollars per million tokens · most-agreed
Input
free
Output
free
Cache read
free
Cache write
free
Quality
8 benchmarks · 2 sources
Composite
85th
percentile of 334 scored models
Rank
51 / 334
±0.08 sd
Evidence
8 × 2
3 or more benchmarks, from 2 or more sources
Effort range
16.5
points between effort settings
reasoning81st3/4 bench
math88th3/7 bench
knowledge73rd1/1 bench
general80th1/1 bench
By reasoning effort
every setting placed on the same scale as the leaderboard
| Setting | Composite | Benchmarks behind it |
|---|---|---|
| high | 91st | 3 |
| max | 89th | 7 |
| low | 69th | 4 |
Every score
one row per source and configuration — nothing averaged away
| Benchmark | Score | Configuration | Source | Run |
|---|---|---|---|---|
| GPQA diamond | 93.1 ±1.5 | max | Epoch AI | 2026-07-16 |
| ↳ | 91.9 ±1.9 | high | Epoch AI | 2026-08-07 |
| ↳ | 84.8 ±2.6 | low | Epoch AI | 2026-08-07 |
| Settings differ by 8.3 points on this benchmark. | ||||
| OTIS Mock AIME 2024-2025 | 97.2 ±1.1 | max | Epoch AI | 2026-07-16 |
| ↳ | 93.3 ±3.8 | high | Epoch AI | 2026-08-07 |
| ↳ | 68.9 ±7.0 | low | Epoch AI | 2026-08-07 |
| Settings differ by 28.3 points on this benchmark. | ||||
| FrontierMath-Tiers-1-3-v2 | 72.2 ±2.7 | max | Epoch AI | 2026-07-17 |
| LiveBench | 79.5 | default · 23 runs | LiveBench | 2026-06-25 |
| Chess Puzzles | 39.0 ±4.9 | max | Epoch AI | 2026-07-16 |
| ↳ | 25.0 ±4.4 | high | Epoch AI | 2026-08-07 |
| ↳ | 20.0 ±4.0 | low | Epoch AI | 2026-08-07 |
| Settings differ by 19.0 points on this benchmark. | ||||
| SimpleQA Verified | 50.6 ±1.6 | max | Epoch AI | 2026-08-27 |
| FrontierMath-Tier-4-v2 | 39.0 ±7.7 | max | Epoch AI | 2026-07-17 |
| Mystery Game Puzzles | 26.0 ±4.4 | max | Epoch AI | 2026-07-29 |
| ↳ | 12.0 ±3.3 | low | Epoch AI | 2026-08-29 |
| Settings differ by 14.0 points on this benchmark. | ||||
The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.
Available from 1 host
| Host | In $/M | Out $/M | Cache rd | Cache wr | Context | Output | Retires |
|---|---|---|---|---|---|---|---|
| kimi-for-coding | free | free | free | free | 1M | 131K | — |
Other listings of this model
this page is one of them
The corpus files Kimi K3 under several keys. This page is the k3 listing. The full record — every host, every price and every benchmark score — is on the main Kimi K3 page.