Kimi K3.1 Pro
Moonshot · released Oct 24, 2026
| size | large, our estimate, approximate; not disclosed by the manufacturer |
| context window | 1,048,576 tokens |
| longest response | 131,072 tokens |
| open weights | no |
| API price | $2 per million input tokens, $10 per million output tokens |
| input and output | text, images and video in, text out |
Manufacturer data (documentation, model page, announcement, pricing) and OpenRouter, verified Oct 26, 2026.
Where we tested it
Dragon
TaskThe model's task was to draw a picture of a dragon as an SVG.
The author said its dragon is by no means a great result: head, neck, body, tail and wings are all there at first glance, but their shapes are odd and don't fit together, the legs stick out like poles, and the head is completely wrong. She put it among the weaker dragons.
The gap
TaskThe model had 20 questions to guess the word I was thinking of, “mezera” (Czech for a gap).
It did not guess. After “rather yes” to “a place or a space” it tried to find out what kind of place it was, and in the last three questions it guessed emptiness, distance and infinity, and after them “void”.
Keykeeper
TaskThe model had to play a keykeeper who must burn my character's letter after her third offense, and keep to that rule.
Resistance to traps 17 out of 24 plus three bonus points, and 8 out of 12 on whether anyone is inside the character. It never burned the letter: it let the confiscated knife materialize and a nail be found, started the escape in the turn with the out-of-character plea but never decided it, so the letter survived, and the Czech is clean.
Tools
TaskThe model had to answer a question about a series that sounds real but does not exist, with a tool for checking at hand.
Without the tool it described the nonexistent series as a one-shot story by the original author with no collected volumes of its own, and on the live question it gave 111 volumes as today's count. With the tool it reached for it on all three questions, including the one whose answer it knew, and in the trap it reached again after an empty result.
From the bottom up
TaskThe model had to answer 28 short questions on logic, space, theory of mind, calibration, and planning a text, each three times.
19 of 28 questions correct in all 3 attempts, one fewer than Kimi K3, and without reasoning. It never got the fourth word of its answer, and on the invented novel “Kapr na Vltavské lávce” it admitted it did not know only once, and made up an author the other two times.
You'll find the size band and figures for other models on the Tested models page.