Glimmer Index
Plus $4/mo

Kimi K3

Moonshot AI · Released Jul 16, 2026 · Open weights

Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work

Strongest at
Writing
Top 13% of the past year's models · 4 of 4 benchmarks
Price per 1M tokens
$3 / $15
input / output · Moonshot AI
Context window
1.05M
tokens in · 131.1K out
Since the last one
87 days
after Kimi K2.6
Lab post ↗Open in the timeline

Capability profile

Scored against 91 models released in the past year. 50 is the average model; further out is better.

This modelKimi K2.6Median of the past year
5090Median of the past yearKimi K2.6Coding: 63 · Top 20%Writing: 64 · Top 13%Reading & documents: 58 · Top 36%Reasoning: 62 · Top 43%Math: 61 · Top 35%Science: 62 · Better than 34%Visual reasoning: 49 · Better than 36%Computer use: 70 · Top 38%CodingWritingReadingReasoningMathScienceVisionComputer use
Coding63+11.1Top 20%
Writing64+8.5Top 13%
Reading & documents58+22.5Top 36%
Reasoning62+16.2Top 43%
Math61+11.1Top 35%
Science62−0.3Better than 34%
Visual reasoning49+15.3Better than 36%
Computer use70+35.7Top 38%
Where these scores come from

Each capability combines the published benchmarks listed here, rescaled so 50 is the reference average and 15 points is one standard deviation. A hollow dot rests on a single benchmark.

Price against capability

The shaded corner holds anything better and cheaper. Prices blend input and output 3 to 1.

4 models from the past year score higher on Writing for less: Kimi K3.1 Pro, Muse Spark 1.3, Gemini 3.8 Flash and 1 more.

$0.1$0.3$1$3$10204060Kimi K3.1 Pro: 70, $4 per 1MClaude Haiku 5.5: 46, $0.2 per 1MMistral Large 4: 41, $2.06 per 1MGemini 4 Argon: 63, $4 per 1MGPT-6.1 Sol: 64, $4 per 1MClaude Sonnet 5.5: 55, $4 per 1MGPT-6 Sol: 61, $4 per 1MGPT-6 Luna: 22, $0.2 per 1MClaude Opus 5.5: 65, $8 per 1MGrok 4.7: 64, $3 per 1MDeepSeek V4.1 Flash: 56, $0.21 per 1MGPT-6 Astra: 69, $20 per 1MMuse Spark 1.3: 65, $2 per 1MGemini 3.8 Flash: 65, $1.5 per 1MClaude Fable 5.1: 69, $20 per 1MGLM-5.3-Flash: 15, $0.24 per 1MDeepSeek V4 Flash Vision Exp: 56, $0.32 per 1MQwen3.8 27B: 52, $1.12 per 1MGLM-5.3: 65, $2.15 per 1MGemini 3.7 Flash: 61, $1.5 per 1MDeepSeek V4 Pro 0813: 49, $0.99 per 1MQwen3.8 2.4T A95B: 62, $3 per 1MGrok 4.6: 44, $3 per 1MMuse Glimmer 30B: 59, $0.52 per 1MMuse Spark 1.2: 62, $2 per 1MQwen3.8 Max: 57, $3 per 1MDeepSeek V4 Flash 0731: 52, $0.15 per 1MClaude Opus 5: 64, $10 per 1MGemini 3.6 Flash: 63, $1.5 per 1MGemini 3.5 Flash Lite: 53, $0.85 per 1MGPT-5.6 Terra: 57, $4.5 per 1MGPT-5.6 Sol: 65, $8 per 1MGPT-5.6 Luna: 53, $0.45 per 1MGrok 4.5: 44, $3 per 1MClaude Sonnet 5: 57, $4 per 1MGLM-5.2: 55, $2.15 per 1MKimi K2.7 Code: 23, $1.71 per 1MClaude Fable 5: 68, $20 per 1MClaude Opus 4.8: 62, $10 per 1MQwen3.7 Max: 48, $3.75 per 1MGemini 3.5 Flash: 56, $3.38 per 1MDeepSeek V4 Pro: 54, $1.19 per 1MDeepSeek V4 Flash: 52, $0.11 per 1MGPT-5.5: 64, $11.2 per 1MQwen3.6 27B: 16, $1.35 per 1MQwen3.6 Max Preview: 44, $2.92 per 1MGrok 4.3: 29, $1.56 per 1MGrok Build 0.1: 43, $1.25 per 1MClaude Opus 4.7: 61, $10 per 1MMuse Spark 1.1: 58, $2 per 1MGLM-5.1: 56, $2.15 per 1MGemma 4 31B IT: 54, $0.25 per 1MGemma 4 26B A4B IT: 51, $0.14 per 1MQwen3.6 Plus: 31, $1.12 per 1MGPT-5.4 nano: 49, $0.46 per 1MGPT-5.4 mini: 52, $1.69 per 1MGPT-5.4: 63, $5.62 per 1MGemini 3.1 Flash Lite Preview: 43, $0.56 per 1MQwen3.5 35B-A3B: 39, $0.69 per 1MQwen3.5 27B: 52, $0.82 per 1MQwen3.5 122B-A10B: 47, $1.1 per 1MGemini 3.1 Pro Preview: 57, $4.5 per 1MClaude Sonnet 4.6: 57, $6 per 1MQwen3.5 397B-A17B: 59, $1.35 per 1MGLM-5: 60, $1.55 per 1MClaude Opus 4.6: 56, $10 per 1MKimi K2.5: 62, $1 per 1MGLM-4.7: 58, $1 per 1MGemini 3 Flash Preview: 49, $1.12 per 1MGPT-5.2 Codex: 46, $4.81 per 1MGPT-5.2: 59, $4.81 per 1MMistral Large 3: 38, $0.75 per 1MMinistral 3 14B: 28, $0.2 per 1MDeepSeek V3.2: 49, $0.32 per 1MKimi K2 Thinking: 61, $1.07 per 1MKimi K2.6: 56, $1.71 per 1MKimi K3: 64, $6 per 1MKimi K3Price per 1M tokens →↑ Writing$0.1$1$10204060Kimi K3.1 Pro: 70, $4 per 1MClaude Haiku 5.5: 46, $0.2 per 1MMistral Large 4: 41, $2.06 per 1MGemini 4 Argon: 63, $4 per 1MGPT-6.1 Sol: 64, $4 per 1MClaude Sonnet 5.5: 55, $4 per 1MGPT-6 Sol: 61, $4 per 1MGPT-6 Luna: 22, $0.2 per 1MClaude Opus 5.5: 65, $8 per 1MGrok 4.7: 64, $3 per 1MDeepSeek V4.1 Flash: 56, $0.21 per 1MGPT-6 Astra: 69, $20 per 1MMuse Spark 1.3: 65, $2 per 1MGemini 3.8 Flash: 65, $1.5 per 1MClaude Fable 5.1: 69, $20 per 1MGLM-5.3-Flash: 15, $0.24 per 1MDeepSeek V4 Flash Vision Exp: 56, $0.32 per 1MQwen3.8 27B: 52, $1.12 per 1MGLM-5.3: 65, $2.15 per 1MGemini 3.7 Flash: 61, $1.5 per 1MDeepSeek V4 Pro 0813: 49, $0.99 per 1MQwen3.8 2.4T A95B: 62, $3 per 1MGrok 4.6: 44, $3 per 1MMuse Glimmer 30B: 59, $0.52 per 1MMuse Spark 1.2: 62, $2 per 1MQwen3.8 Max: 57, $3 per 1MDeepSeek V4 Flash 0731: 52, $0.15 per 1MClaude Opus 5: 64, $10 per 1MGemini 3.6 Flash: 63, $1.5 per 1MGemini 3.5 Flash Lite: 53, $0.85 per 1MGPT-5.6 Terra: 57, $4.5 per 1MGPT-5.6 Sol: 65, $8 per 1MGPT-5.6 Luna: 53, $0.45 per 1MGrok 4.5: 44, $3 per 1MClaude Sonnet 5: 57, $4 per 1MGLM-5.2: 55, $2.15 per 1MKimi K2.7 Code: 23, $1.71 per 1MClaude Fable 5: 68, $20 per 1MClaude Opus 4.8: 62, $10 per 1MQwen3.7 Max: 48, $3.75 per 1MGemini 3.5 Flash: 56, $3.38 per 1MDeepSeek V4 Pro: 54, $1.19 per 1MDeepSeek V4 Flash: 52, $0.11 per 1MGPT-5.5: 64, $11.2 per 1MQwen3.6 27B: 16, $1.35 per 1MQwen3.6 Max Preview: 44, $2.92 per 1MGrok 4.3: 29, $1.56 per 1MGrok Build 0.1: 43, $1.25 per 1MClaude Opus 4.7: 61, $10 per 1MMuse Spark 1.1: 58, $2 per 1MGLM-5.1: 56, $2.15 per 1MGemma 4 31B IT: 54, $0.25 per 1MGemma 4 26B A4B IT: 51, $0.14 per 1MQwen3.6 Plus: 31, $1.12 per 1MGPT-5.4 nano: 49, $0.46 per 1MGPT-5.4 mini: 52, $1.69 per 1MGPT-5.4: 63, $5.62 per 1MGemini 3.1 Flash Lite Preview: 43, $0.56 per 1MQwen3.5 35B-A3B: 39, $0.69 per 1MQwen3.5 27B: 52, $0.82 per 1MQwen3.5 122B-A10B: 47, $1.1 per 1MGemini 3.1 Pro Preview: 57, $4.5 per 1MClaude Sonnet 4.6: 57, $6 per 1MQwen3.5 397B-A17B: 59, $1.35 per 1MGLM-5: 60, $1.55 per 1MClaude Opus 4.6: 56, $10 per 1MKimi K2.5: 62, $1 per 1MGLM-4.7: 58, $1 per 1MGemini 3 Flash Preview: 49, $1.12 per 1MGPT-5.2 Codex: 46, $4.81 per 1MGPT-5.2: 59, $4.81 per 1MMistral Large 3: 38, $0.75 per 1MMinistral 3 14B: 28, $0.2 per 1MDeepSeek V3.2: 49, $0.32 per 1MKimi K2 Thinking: 61, $1.07 per 1MKimi K2.6: 56, $1.71 per 1MKimi K3: 64, $6 per 1MKimi K3Price per 1M tokens →↑ Writing

Moonshot AI $3 / $15 · OpenRouter $0.17 / $20 · Cached input $0.3

Since Kimi K2.6

5 releases in the Kimi line.

Kimi K2 0711Jul 11, 2025200dKimi K2.5Jan 27, 202683dKimi K2.6Apr 20, 202687dKimi K3Jul 16, 202657dKimi K2.8 PreviewSep 11, 2026
Released87 days later
Price in / out$3 / $15+250%
Context1.05M4.0×
Coding63+11.1
Writing64+8.5
Reading & documents58+22.5
Reasoning62+16.2
Math61+11.1
Science62−0.3
Visual reasoning49+15.3
Computer use70+35.7

A newer release is out: Kimi K2.8 Preview.

ECI STANDING#19 of 270
SCORE157.86155.11–160.63 CI
FRONTIER LAGLaunched 4.4 months behindScores about what GPT-5.4 Pro did in March 2026.
SCORES LIKE THESE MODELSGemini 3.7 FlashGoogle DeepMind · Aug 2026157.97Gemini 3.8 FlashGoogle DeepMind · Sep 2026157.44Qwen 3.8 MaxAlibaba · Aug 2026157.4GPT-5.4OpenAI · Mar 2026157.31
Full Epoch Capabilities Index ranking ↗

Epoch benchmark boards

Best result on each board it has run. The tick marks the leader.

Surface Evolver Bench
95%#2 of 26
WeirdML
83%#11 of 141
GPQA diamond
93%#21 of 234
ProofBench
87%#10 of 81
OTIS Mock AIME 2024-2025
97%#35 of 216
ARC-AGI
94%#18 of 107
DTBench
91%#37 of 179
FrontierMath-Tiers-1-3-v2-Private
72%#25 of 90
12 more boards
SimpleBench
61%#26 of 89
ARC-AGI-2
60%#32 of 103
SimpleQA Verified
51%#27 of 85
FrontierCode
44%#16 of 46
FrontierMath-Tier-4-v2-Private
39%#28 of 72
Mystery Game Puzzles
26%#35 of 88
Chess Puzzles
37%#40 of 93
PostTrainBench
32%#5 of 11
APEX-Agents
48%#28 of 51
FrontierSWE
26%#12 of 21
WeirdML v3
15%#14 of 23
Furniture Assembly
34%#23 of 35

Moonshot AI's past year

7 release days in the past year, about one every 52 days. The next is expected around Dec 10, 2026. Forecasts ↗

NovDecJanFebMarAprMayJunJulAugSepOctNovDec

Specifications

RELEASE DATEJul 16, 2026
WEIGHTSOpen weights
CONTEXT WINDOW1.05M
MAX OUTPUT131.1K
REASONINGNot listed
TOOL CALLINGNot listed
INPUTStext, image/jpeg, image/png, image/gif, image/webp, image/bmp, image/heic, image/heif, video/mp4, video/mpeg, video/mov, video/avi, video/x-flv, video/mpg, video/webm, video/wmv, video/3gpp
OUTPUTStext

Models.dev record ↗

Specifications and prices from Models.dev; availability and limits may differ by provider. Benchmarks from Epoch AI (CC BY 4.0), LiveBench and the other sources named above.