ZMIME
Moonshot AI · Released Jul 16, 2026

Kimi K3

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

  • Open weights
  • Reasoning
  • Tool calling
  • Vision
  • Video input
Capability
158.6ECI · #25 of 158
Input
$3.00per 1M tokens
Output
$15.00per 1M tokens
Context
1.05M1.05M max output
Pricing

What it costs

Input
$3.00per million tokens
Output
$15.00per million tokens
Cached input
$0.30per million tokens
Cache write
$3.00per million tokens
Blended (3:1)
$6.00pricier than 91% of priced models

Typical monthly bills

Monthly usageEstimated cost
Side project2M input + 0.5M output tokens$13.50
Team assistant25M input + 5M output tokens$150.00
Production app250M input + 50M output tokens$1,500

Official Moonshot AI API price, as listed on models.dev.

Capability

Independent benchmarks

158.6Epoch Capabilities Index
#25 of 158 scored models
68Median 147174

The shaded band is Epoch AI’s confidence range (156.6–160.8); the tick marks the median scored model.

  • GPQA DiamondGraduate-level science questions · max setting93.1%
  • FrontierMath Tiers 1–3Research-level mathematics · max setting72.2%
  • OTIS Mock AIME 2024–2025Competition mathematics · max setting97.2%
  • SimpleQA VerifiedShort factual questions · max setting50.6%

Scores from Epoch AI, run independently of Moonshot AI.

Specifications

The details

Lab
Moonshot AI
Released
Jul 16, 2026
Context window
1,048,576 tokens
Max output
1,048,576 tokens
Inputs
Text, Images, Video
Output
Text
Reasoning
Adjustable effort low · high · max
Tool calling
Yes
Structured output
Yes
Weights
Open weights
API model ID
kimi-k3on Moonshot AI
Availability
94 API providerslisted on models.dev

Moonshot AI model documentation

Lab-reported

What Moonshot AI claims

Published by the lab at launch. Settings vary, so compare these only with care.

BenchmarkScoreSettingSource
DeepSWE v1.167.5 resolve ratemax effortSource
Terminal-Bench v2.188.3max effortSource
FrontierSWE81.2max effortSource
Program Bench77.8max effortSource
SWE Marathon v1.142 resolve ratemax effortSource
GDPval-AA vv21668 Elomax effortSource
AA-Briefcase1548 Elomax effortSource
AutomationBench30.8 success ratemax effortSource
JobBench52.9max effortSource
SpreadsheetBench v234.8max effortSource
BrowseComp91.2max effort, context compactionSource
CharXiv Reasoning91.3max effort, with toolsSource
Same lab

More from Moonshot AI

Questions

About Kimi K3

How much does Kimi K3 cost?

Kimi K3 costs $3.00 input / $15.00 output per million tokens (official Moonshot AI API price). Cached input is $0.30 per million tokens, and writing to the cache costs $3.00. At a 3:1 input-to-output mix that is $6.00 per million tokens, more expensive than 91% of the 469 priced models we track.

What is the context window of Kimi K3?

Kimi K3 accepts up to 1,048,576 tokens per request and can write up to 1,048,576 tokens in one response.

How good is Kimi K3?

Epoch AI gives Kimi K3 a Capabilities Index score of 158.6 (likely range 156.6–160.8), ranking it #25 of 158 models Epoch has scored. Epoch AI benchmark results: GPQA Diamond 93.1%, FrontierMath Tiers 1–3 72.2%, OTIS Mock AIME 2024–2025 97.2%, SimpleQA Verified 50.6%.

Is Kimi K3 open source?

Yes. Moonshot AI publishes the weights, so it can be downloaded and self-hosted.

What inputs does Kimi K3 support?

Kimi K3 accepts text, images and video and replies in text. It is a reasoning model with low, high and max effort settings, supports tool calling and can return structured JSON output.

When was Kimi K3 released?

Moonshot AI released Kimi K3 on Jul 16, 2026.

What are the best alternatives to Kimi K3?

The closest current models from other labs on capability, price and release date are GPT-5.4 (OpenAI, ECI 157.9, $2.50 / $15.00), Claude Sonnet 5 (Anthropic, ECI 156.9, $2.00 / $10.00), Grok 4.6 (xAI, ECI 157.8, $2.00 / $6.00) and Qwen3.8 Max (Alibaba (Qwen), ECI 157.6, $2.00 / $6.00).