NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Kimi K3
moonshotai logo

Kimi K3

by moonshotai · Released Jul 2026

Multimodal1M Context
64.5
avg score
Rank #56
Compare
Better than 83% of all models
Context
1.0M tokens (~524 books)
Input $/1M
$0.19
Output $/1M
$16.00
Type
multimodal
License
Proprietary
Benchmarks
31 tested
Data as of
About

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...

Tested on 31 benchmarks · BenchGecko score 64.5. Top scores: OTIS Mock AIME 2024-2025 (97.2%), Surface Evolver Bench (95.0%), GPQA diamond (94.6%).

Looking for similar performance at lower cost?
GLM 5.2 scores 64.4 (100% as good) at $0.01/1M input · 96% cheaper
Capabilities
coding
55.1
#68 globally
reasoning
69.3
#56 globally
math
70.7
#50 globally
knowledge
45.6
#174 globally
agentic
47.8
#30 globally
speed
61.0
#8 globally
general
59.0
#22 globally
Benchmark Scores
Compare All
Tested on 31 benchmarks · Ranked across 7 categories
Score Distribution (all 326 models)
0255075100
▲ You are here
codingCompare coding →
WeirdML

Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.

82.6·
Deepswe
68.5·
Mirrorcode
54.4·
reasoningCompare reasoning →
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

94.5·
ARC-AGI-2

ARC-AGI 2, harder sequel to ARC. More complex abstract reasoning patterns that test generalization ability beyond training data.

60.4·
SimpleBench

Deceptively simple questions that humans find easy but AI models often get wrong. Tests common sense and reasoning gaps.

52.8·
mathCompare math →
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

97.2·
FrontierMath-Tiers-1-3-v2-Private
76.0·
FrontierMath-Tier-4-v2-Private
39.0·
Quick compare:z-ai logovs GLM 5.2Anthropic logovs Claude Haiku 5.5Google DeepMind logovs Gemini 3.7 Flash
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Recently Happened
Kimi K3 pricing increased 7%
Nov 4, 2026
Kimi K3 pricing dropped 25%
Oct 30, 2026
Kimi K3 pricing dropped 21%
Oct 19, 2026
Kimi K3 pricing dropped 53%
Oct 15, 2026
Kimi K3 pricing increased 239%
Sep 26, 2026
Kimi K3 pricing increased 76%
Sep 22, 2026
Similar Models
z-ai logo
GLM 5.2
z-ai
64.4$0.01/1M
Anthropic logo
Claude Haiku 5.5
Anthropic
64.4$0.10/1M
Google DeepMind logo
Gemini 3.7 Flash
Google DeepMind
64.3$0.75/1M
Links
Info
moonshotaiPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@moonshotai
BenchGecko API
kimi-k3
Specifications
  • Typemultimodal
  • Context1.0M tokens (~524 books)
  • ReleasedJul 2026
  • LicenseProprietary
  • StatusActive
  • Cost / Message~$0.016
Available On
moonshotai logomoonshotai$0.19
Categories
codingreasoningmathknowledgeagenticspeedgeneral
Learn More
context-windowtransformertokens
Share & Export
Tweet
Related Models
z-ai logoGLM 5.2
64.4
Anthropic logoClaude Haiku 5.5
64.4
Google DeepMind logoGemini 3.7 Flash
64.3
OpenAI logoGPT-5.4
64.3
Kimi K3 is a proprietary multimodal AI model by moonshotai, released in July 2026. It has an average benchmark score of 64.5. Context window: 1M tokens.

Related Models

GLM 5.2 · z-aiClaude Haiku 5.5 · AnthropicGemini 3.7 Flash · Google DeepMindGPT-5.4 · OpenAIFalcon 2 11B · TII

Benchmarks

OTIS Mock AIME 2024-2025Surface Evolver BenchGPQA diamondARC-AGIArtificial Analysis · GPQA Diamond

Related Pages

moonshotai · Providermoonshotai · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-11-11

  • Kimi K3 by moonshotai. BenchGecko score 64.5, rank 56 of 326 scored models (normalized average of public benchmark scores).
  • List price $0.19 input · $16.00 output per 1M tokens (as of 2026-11-11).
  • Sold by 28 providers (as of 2026-11-11): Wafer (mxfp4) $0.15 in / $16.00 out · Relace (fp4) $0.18 in / $11.01 out · Wafer (mxfp4) $0.19 in / $16.00 out · InferenceNet (fp4) $0.23 in / $15.00 out · InferenceNet (fp4) $0.23 in / $15.00 out · and 23 more. Every provider
  • Gecko Tests · Gecko Score 40 (rank 40): Who Are You B (Knows who made it) · World Map C (94.6% of the map right) · Knowledge Horizon C (Knows news up to Nov 2025) · Tokenizer Tax E (93% more tokens outside English).

How to cite · data as of 2026-11-11

Kimi K3 · benchmarks, pricing and providers. BenchGecko, data as of 2026-11-11. https://benchgecko.ai/model/kimi-k3

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP