Moonshot AI

Kimi K3

Version 3 benchmark runs across the shared Superbash visual prompts.

API pricing

USD per million tokens · Kimi API
Input
$3
Output
$15
Cached input
$0.3
Context window
1.05M
Pricing details and specifications

Suggested for this guide

Kimi

Use the same Kimi family tested on this page. Check the current plan and model access before subscribing.

Best for: Claude Code setups and cost-aware model work

Check Kimi plansPartner link. It supports Superbash Learn at no extra cost to you.

Inspect the original visual runs

Open the generated scene to test it, or compare that prompt with another model.

4 runs

Helm's Deep

Run date: 2026-07-16

Generation provenance: unrecorded

Fortress siege scene testing scale, lighting, architecture, and cinematic atmosphere.

Hogwarts Broom Flight Simulator

Run date: 2026-07-16

Generation provenance: unrecorded

Broom-flight scene testing depth, motion cues, castle scale, and fantasy mood.

Mechanical Watch Simulator

Run date: 2026-07-24

Generation provenance: unrecorded

Interactive watch movement testing mechanical legibility, accurate relative motion, and real-time 3D controls.

Yingzao Fashi Assembly

Run date: 2026-07-16

Specialist test · evaluated separately

Generation provenance: unrecorded

Timber assembly scene testing structure, joinery, construction order, and material clarity.

Retired tests · 6 archived runs

These tests are no longer in the active suite. Their original results remain available for inspection.

How Kimi K3 compares

Three relevant peers, side by side.

  • Kimi K3 This model
  • Claude Fable 5
  • GPT-5.6 Sol
  • Claude Opus 4.8

6 benchmarks · 4 models

GPQA Diamond

Academic reasoning · Higher is better

Source
Kimi K393.5%
Claude Fable 592.6%
GPT-5.6 Sol94.1%
Claude Opus 4.8Not reported

DeepSWE

Coding · Higher is better

Source
Kimi K367.5%
Claude Fable 570.0%
GPT-5.6 Sol73.0%
Claude Opus 4.859.0%

Terminal-Bench 2.1

Agentic coding · Higher is better

Source
Kimi K388.3%
Claude Fable 588.0%
GPT-5.6 Sol88.8%
Claude Opus 4.884.6%

SWE-Marathon

Long-horizon coding · Higher is better

Source
Kimi K342.0%
Claude Fable 535.0%
GPT-5.6 Sol39.0%
Claude Opus 4.840.0%

BrowseComp

Agentic research · Higher is better

Source
Kimi K391.2%
Claude Fable 588.0%
GPT-5.6 Sol90.4%
Claude Opus 4.8Not reported

DeepSearchQA (F1)

Deep research · Higher is better

Source
Kimi K395.0%
Claude Fable 594.2%
GPT-5.6 SolNot reported
Claude Opus 4.893.1%

Official benchmark profile

Published scores outside our visual tests

Moonshot AI sourceAugust 2026Source report →
Academic reasoning

GPQA Diamond

93.5%
Coding

DeepSWE

67.5%
Agentic coding

Terminal-Bench 2.1

88.3%
Full official benchmark table12 scores and peer comparisons
BenchmarkAreaScoreComparison
GPQA DiamondAcademic reasoning93.5%
Kimi K393.5%
GPT-5.6 Sol94.1%
Claude Fable 592.6%
GLM 5.291.2%
DeepSWECoding67.5%
Kimi K367.5%
GPT-5.567.0%
Claude Fable 570.0%
GPT-5.6 Sol73.0%
Terminal-Bench 2.1Agentic coding88.3%
Kimi K388.3%
Claude Fable 588.0%
GPT-5.6 Sol88.8%
Claude Opus 4.884.6%
SWE-MarathonLong-horizon coding42.0%
Kimi K342.0%
Claude Opus 4.840.0%
GPT-5.6 Sol39.0%
Claude Fable 535.0%
BrowseCompAgentic research91.2%
Kimi K391.2%
GPT-5.6 Sol90.4%
Claude Fable 588.0%
GPT-5.584.4%
DeepSearchQA (F1)Deep research95.0%
Kimi K395.0%
Claude Fable 594.2%
Claude Opus 4.893.1%
MCPMark-VerifiedTool use94.5%
Kimi K394.5%
GPT-5.6 Sol92.9%
GPT-5.592.9%
Claude Fable 587.4%
AutomationBenchWorkflow automation30.8%
Kimi K330.8%
GPT-5.6 Sol29.7%
Claude Fable 529.1%
Claude Opus 4.827.2%
SpreadsheetBench 2Spreadsheet work34.8%
Kimi K334.8%
Claude Fable 534.7%
GPT-5.6 Sol32.4%
Claude Opus 4.831.6%
OmniDocBenchDocument understanding91.1%
Kimi K391.1%
Claude Fable 589.8%
GPT-5.589.4%
Claude Opus 4.887.9%
MMMU-ProMultimodal reasoning81.6% / 83.4%
Kimi K381.6% / 83.4%
Claude Fable 581.2% / 86.5%
GPT-5.581.2% / 83.2%
GPT-5.6 Sol83.0% / 84.6%
CharXiv (RQ)Chart reasoning84.8% / 91.3%
Kimi K384.8% / 91.3%
GPT-5.6 Sol84.6% / 89.1%
GPT-5.584.1% / 89.0%
Claude Fable 588.9% / 93.5%
Sources, pricing details and methodology

Verified model facts

Model specifications

Pricing source · checked 2026-09-08 ↗
Status
Superseded in this ranking
API model ID
kimi-k3
Context
1.05M
Max output
131K
License
Not publicly verified
Access route
Not publicly verified
Modalities
Not publicly verified
Hardware
Not publicly verified
Kimi API · USD / 1M tokens
$3 input · $0.3 cached input · $15 output

Prices exclude applicable taxes and separately billed tools.

Benchmark sources

Moonshot reports Kimi K3 at maximum reasoning effort across coding, research, tool use, document work, and vision. The official table mixes provider harnesses on some agent tasks, so each row keeps its evaluation setting.

Kimi K3 Technical Report

Evaluation settings

  • Maximum reasoning effort, temperature 1.0, top-p 0.95.
  • Kimi Code harness on DeepSWE v1.1 tasks.
  • Kimi Code harness; peers use each provider's cited best harness.
  • Claude Code for Kimi and Claude; Codex for GPT-5.6 on the July 9 H20-calibrated task branch.
  • Context compaction at 300K tokens; the full 1M-context run without context management scored 90.4%.
  • Maximum reasoning effort, temperature 1.0, top-p 1.0.
  • 600-task public subset with the official benchmark setup.
  • Claude Code for Kimi and Claude; Codex for GPT models.
  • Maximum reasoning effort; average over three multimodal runs.
  • Without tools / with Python; original input order; average over three runs.
  • Without tools / with Python; average over three runs.
DeepSWE
The official leaderboard score under mini-SWE-agent is 67.3%.