Skip to content

Kimi K3

ModelOpen sourceActive
by Moonshot AI · family “kimi-k”

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... (description from the OpenRouter listing)

in: textin: imagein: videoout: textReasoningTool useStructured output
Epoch Capabilities Index
157.7 #12 of 268
90% CI 155.1 – 160.7
Listing ↗
Released
16 Jul 2026
Context window
1.05M tokens
Max output
944K tokens
Input $ / 1M tokens
$3.00
Output $ / 1M tokens
$15.0
Cached input $ / 1M
$0.30
Each benchmark shown separately with its own source

Benchmark results (16)

All benchmarks
BenchmarkDomainScorevs best recordedSettingRunSource
Furniture Assemblymultimodal34.2% ±6.1
41%
max11 Sep 2026Epoch ↗
LMCAagents52.7%
77%
max—External ↗
DTBenchreasoning91.2%
92%
max—External ↗
Mystery Game Puzzlesgames26.0% ±4.4
31%
max29 Jul 2026Epoch ↗
Surface Evolver Benchscience93.0%
98%
max—External ↗
FrontierMath-Tiers-1-3-v2-Privatemath72.2% ±2.7
77%
max17 Jul 2026Epoch ↗
FrontierMath-Tier-4-v2-Privatemath39.0% ±7.7
40%
max17 Jul 2026Eval log ↗
DeepSWEcoding68.5%
92%
max—External ↗
Chess Puzzlesgames39.0% ±4.9
54%
max16 Jul 2026Epoch ↗
SimpleQA Verifiedknowledge50.6% ±1.6
67%
max27 Aug 2026Eval log ↗
ARC-AGI-2reasoning60.4%
64%
max—External ↗
WeirdMLcoding82.6% ±0.0
88%
max—External ↗
OTIS Mock AIME 2024-2025math97.2% ±1.1
97%
max16 Jul 2026Epoch ↗
SimpleBenchreasoning60.7%
74%
max—External ↗
GPQA diamondscience93.1% ±1.5
97%
max16 Jul 2026Epoch ↗
ARC-AGIreasoning94.5%
96%
max—External ↗

Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.

API list price over time

Price history

$ per 1M tokens

4 recorded prices on 17 Jul 2026 (OpenRouter listing + Internet Archive snapshots); re-read on every data refresh, most recently 10h ago. Steps show when the price changed.

Events