Skip to content
Leaderboard / Moonshot AI

Kimi K3

Moonshot AI Max OPEN Kimi K3 license
Verdict

Kimi K3 is a strong all-rounder: #35 of 406 by SIscope score. Strongest at following instructions, weaker at agents & tools. Mid-range on price. Weights are open, so you can run it yourself.

SIscope score
82
#35 of 406
Human-vote rank
#25
of 419 · score 1476
Price per 1M tokens, blended
$2.61
$0.28 in · $9.6 out
Context window
1M
1,048,576 tokens

Shape

Each spoke is a category, 0 at the centre, 100 at the edge.

CodingAgentsReasoningMathChatInstructions

Category scores

Percentiles: 90 means better than about 90% of tested models.

Coding92
Agents & tools57
Reasoning & knowledge88
Math87
Chat quality94
Following instructions96

What goes into each category · Compare with Gemini 4 Carbon

Human-vote score over time

Text arena, recent updates.

147014751480Oct 15: 1474.7 (rank #25)Oct 20: 1475.7 (rank #24)Oct 22: 1475 (rank #25)Oct 28: 1475.7 (rank #25)Oct 15Oct 28

Coding votes over time

WebDev arena, recent updates.

165016581665Oct 16: 1655.2 (rank #14)Oct 18: 1654.6 (rank #15)Oct 19: 1655.3 (rank #16)Oct 22: 1654.7 (rank #16)Oct 23: 1654.9 (rank #17)Oct 27: 1656.2 (rank #17)Oct 28: 1656.4 (rank #18)Oct 16Oct 28

Independent test runs

Run by Epoch AI on the same harness for every model. Bars show the share of questions answered correctly.

GPQA Diamond93%
PhD-level science multiple choice questions
Better than 93% of 333 tested models. Run Jul 16, 2026.
SimpleQA Verified51%
Short factual questions; rewards not making things up
Better than 74% of 100 tested models. Run Aug 27, 2026.
OTIS Mock AIME 2024-2597%
Competition maths problems in the style of AIME
Better than 87% of 313 tested models. Run Jul 16, 2026.
FrontierMath Tiers 1-372%
Very hard, unpublished research-style maths problems
Better than 79% of 122 tested models. Run Jul 17, 2026.

Human-vote arenas

People compare two anonymous answers and pick the better one. Scores work like chess ratings.

ArenaRankScoreRange (95%)VotesUpdated
Text arena
Everyday chat, blind human votes
#25 of 41914761471–148036,780Oct 28, 2026
WebDev arena
Building web apps, blind human votes
#18 of 15216561651–166221,281Oct 28, 2026
Vision arena
Questions about images, blind human votes
Not on this board yet
Agent arena
Real multi-step agent sessions
#25 of 570.9%0.3–1.4%161,000Nov 2, 2026

Source: Arena (formerly LMArena) leaderboard dataset, CC BY 4.0. Fetched Nov 4, 2026.

Price

Input
$0.28
per 1M tokens
Output
$9.6
per 1M tokens
Blended
$2.61
3 in : 1 out, per 1M
Context window
1M
max tokens per request
Speed
Coming soon
our own tests

List price from the OpenRouter models API (moonshotai/kimi-k3), Nov 4, 2026. Providers can charge differently.