whatmodel.sh / Kimi K3.1 Pro
Kimi K3.1 Pro
↗ share
+ contribute
Kimi K3.1 Pro is a closed-weights language model from Moonshot AI, released October 2026.
Catalogue identity notes
API available; full weights promised by November 4, 2026. Not marked open weights until actually released. Highspeed is not a separate quality identity.
Model information sources
Score within the LMArena (text) comparison set only.
$ how it compares
The LLMs ranked either side of Kimi K3.1 Pro, out of 1192 with a score, then more from Moonshot AI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 12 |  GPT-5.5 Pro | 80±1 | api $67.50/M | ||
| 13 | GPT-5.6 Sol | 80±1 | api $8.00/M | ||
| 14 | Kimi K3.1 Pro | 80±2 | api price unknown | ||
| 15 |  Claude Fable 5 | 79±1 | api $20.00/M | ||
| 16 | Claude Mythos Preview (limited preview) | 79±1 | limited preview | ||
| more from Moonshot AI | |||||
| 32 | Kimi K3 | 75±1 | api $6.00/M +local | ||
| 76 | Kimi K2.6 | 72±1 | api $1.71/M +local | ||
| 90 | Kimi K2.7 Code | 71±1 | api $1.71/M +local | ||
$ what it's good at
agent tasksnot ranked on this task yet · 4 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Humanity's Last Exam with tools lab | 70.0% ✎ | ▂▃▁▂▃▃▃█▇▇▆▃ beats 58/58 | — |
| Terminal-Bench 4.0 lab | 65.7% ✎ | ▁·▁··▁▆▆▅██▃ beats 31/33 | — |
| Agents' Last Exam (metric/harness unspecified) lab | 43.2% ✎ | ▂·▂·█▄▂·▃▂▄▄ beats 9/14 | — |
| Chartography with tools lab | 90.2% ✎ | ▂········▄██ beats 9/10 | — |
coding▁▁▁▃▅▇█▇▅▄▃▃ #9 of 481 * · 98% of its score · 1 test
Beats GPT-6.1 Terra, Gemini 4 Argon, GPT-6 Sol here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena WebDev (Code Arena) | 1694.6 ✎ | ▁▁▃▂▅█▃▆▅▅▃▂ beats 122/129 | 98% |
web searchnot ranked on this task yet · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Humanity's Last Exam with tools lab | 70.0% ✎ | ▂▃▁▂▃▃▃█▇▇▆▃ beats 58/58 | — |
other tests1 not in a task · 0.0% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| ZeroBench main with tools (pass@5) lab | 63.0% ✎ | ▄········█·▄ beats 3/3 | — |
$ where to run it
Through an API
api · price unknown · 1M context · platform.kimi.ai ↗
$ how the score was worked out · LMArena leaderboard dataset 98%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 LMArena leaderboard dataset | 1 | 98% |
| prior · pull toward the middle | 2% | |
| weighted average | = 79.6 |
not counted (5)
| Moonshot AI vendor-reported benchmark scores for its own Kimi models (Hugging Face model cards) | Terminal-Bench 4.0 [correct] | 65.7% | fewer than three models have this test yet, or this source ran it for this model alone |
| Moonshot AI vendor-reported benchmark scores for its own Kimi models (Hugging Face model cards) | Humanity's Last Exam with tools [correct] | 70.0% | fewer than three models have this test yet, or this source ran it for this model alone |
| Moonshot AI vendor-reported benchmark scores for its own Kimi models (Hugging Face model cards) | Chartography with tools [correct] | 90.2% | fewer than three models have this test yet, or this source ran it for this model alone |
| Moonshot AI vendor-reported benchmark scores for its own Kimi models (Hugging Face model cards) | ZeroBench main with tools (pass@5) [correct] | 63.0% | fewer than three models have this test yet, or this source ran it for this model alone |
| Moonshot AI vendor-reported benchmark scores for its own Kimi models (Hugging Face model cards) | Agents' Last Exam (metric/harness unspecified) [correct] | 43.2% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json