Kimi K3
Released 16 Jul 2026$3 input, $15 output per million tokens1M context2.8T total / 104B activeKimi K3 License (open weights)
8 results on 8 of 14 boards, latest measured Oct 2026. Compare with other models
On the Clinical Benchmarks Index
The top 15 on the index, then Kimi K3 at rank 18. Select a column to open that model.
Board by board
Kimi K3 against the top 10 on each board it has a result on, on that board's own scale. Its own column is in colour.
HealthBench Professionalrank 28 of 43
- 0.703GPT-6 Astra (Anthropic run)
- 0.692Claude Sonnet 5.5
- 0.660Claude Fable 5
- 0.656Claude Opus 5.5
- 0.655GPT-6 Luna (Anthropic run)
- 0.648Claude Haiku 5.5
- 0.642GPT-6.1 Sol
- 0.627GPT-6.1 Terra
- 0.623Muse Wonder Full
- 0.532Kimi K3
37 modelsFull board
MAST (Medical AI Superintelligence Test)rank 2 of 8
- 60.2GPT-5.6 Sol
- 60.1Kimi K3
- 59.3Gemini 3.6 Flash
- 58.9Gemini 3.1 Pro
- 57.9Qwen3.5 397B A17B
- 57.1Claude Opus 5
- 56.6Claude Sonnet 5
- 53.7Grok 4.3
8 modelsFull board
Bedside Benchrank 6 of 15
- 94.6Claude Opus 5
- 94.3Muse Spark Open
- 93.4GPT-6 Astra
- 92.7Gemini 4 Argon
- 91.1GPT-5.6 Sol
- 89.7Kimi K3
- 89.1Ember-1
- 88.1GPT-6 Sol
- 87.2GPT-6 Luna
- 86.9GLM 5.3
15 modelsFull board
APEX-1 GPrank 4 of 12
- 73.04Claude Opus 5.5
- 72.12Claude Opus 5
- 68.36Gemini 3.8 Flash
- 64.35Kimi K3
- 63.75GPT-5.6 Sol
- 63.2GLM 5.3
- 61.03GPT-6 Astra
- 60.61GPT-6.1 Sol
- 59.38Grok 4.6
- 59.16Grok 4.7
12 modelsFull board
MedScribe (Vals AI)rank 13 of 106
- 91.43Claude Opus 5.5
- 91.29Claude Fable 5.1
- 91.10Claude Sonnet 5.5
- 90.98Claude Opus 5
- 90.06Muse Spark 1.2
- 89.38Grok 4.7
- 88.94GLM 5.3 Flash
- 88.89Muse Spark 1.1
- 88.81GLM 5.3
- 87.96Kimi K3
95 modelsFull board
MedCode (Vals AI)rank 25 of 104
- 63.57Claude Opus 5
- 59.06Gemini 3.1 Pro Preview (02/26)
- 58.80Gemini 4 Argon
- 56.07Claude Fable 5
- 55.92Gemini 3 Flash (12/25)
- 55.83Gemini 3.5 Flash
- 54.86Claude Opus 4.7
- 53.51Claude Fable 5.1
- 53.39Gemini 3.7 Flash
- 48.88Kimi K3
95 modelsFull board
CHI-Benchrank 7 of 43
- 54.7erius + claude-opus-5
- 37.3erius + claude-opus-4-8
- 28.0claude-code + claude-opus-4-6
- 26.2claude-code + claude-sonnet-4-6
- 25.3codex + gpt-5.6-sol
- 25.3openai-agents + kimi-k3
- 24.4claude-code + claude-opus-4-7
- 24.0claude-code + claude-fable-5
- 20.9codex + gpt-5.5
- 20.0claude-code + claude-sonnet-5
24 modelsFull board
First, Do NOHARM (v2)rank 7 of 17
- 79.7Muse Spark 1.1
- 74.6Claude Opus 5
- 74.0Kimi K3
- 70.1GPT-5.6 Sol
- 70.0GPT-5.5
- 68.6GPT-5
- 65.0Claude Fable 5
- 62.6Gemini 3.1 Pro
- 61.9Gemini 2.5 Pro
- 61.1Qwen3.5 397B A17B
13 modelsFull board
Every result, with its source
Each line sits on its own board's scale. Tall tick: the board leader.
Clinical reasoning and knowledge
- HealthBench ProfessionalArtificial Analysis independent run, max reasoning; GPT-5.4 (low) judge, 4 repeats per task; length-adjusted (56.9 raw, 3,250 mean response chars)0.532Rank 28 of 43Leader GPT-6 Astra (Anthropic run) 0.703
- 60.1Rank 2 of 8 here, 11 models on the boardLeader GPT-5.6 Sol 60.2
- Bedside BenchScore (avg@3)89.7Rank 6 of 15 here, 16 models on the boardLeader Claude Opus 5 94.6
- APEX-1 GPScore (avg@4), reasoning max64.35Rank 4 of 12Leader Claude Opus 5.5 73.04
Documentation and coding
- 87.96Rank 13 of 106 here, 105 models on the boardLeader Claude Opus 5.5 91.43
- MedCode (Vals AI)model ID kimi/kimi-k3; temperature=1; max_output_tokens=3000048.88Rank 25 of 104 here, 103 models on the boardLeader Claude Opus 5 63.57
EHR and workflow agents
- CHI-BenchListed as openai-agents + kimi-k325.3Rank 7 of 43 here, 45 models on the boardLeader erius + claude-opus-5 54.7
Safety
- 74.0Rank 7 of 17 here, 19 models on the boardLeader LiSA 2.5 86.2
Index working
Show the working: each board’s result, how hard the board runs and the field average
The Clinical Benchmarks Index is this site's own calculation from 12 boards. How it is computed
- HealthBench Professional0.53253.2+5.0 easy board48.20.743 weightArtificial Analysis independent run, max reasoning; GPT-5.4 (low) judge, 4 repeats per task; length-adjusted (56.9 raw, 3,250 mean response chars)
- MedScribe (Vals AI)87.9688.0+33.8 easy board54.20.15 weight, near its ceilingNear its ceiling: leader 91.4, weight 0.15
- MedCode (Vals AI)48.8848.9−3.3 hard board52.20.319 weightmodel ID kimi/kimi-k3; temperature=1; max_output_tokens=30000
- MAST (Medical AI Superintelligence Test)60.160.1+8.8 easy board51.30.597 weight
- First, Do NOHARM (v2)74.074.0+18.2 easy board55.80.507 weight
- Bedside Bench89.789.7+39.2 easy board50.50.25 weight, near its ceilingNear its ceiling: leader 94.6, weight 0.25Score (avg@3)
- APEX-1 GP64.3564.4+11.9 easy board52.50.674 weightScore (avg@4), reasoning max
- Adjusted average of 7 boards, by weight
- 51.7
- Field average, counts for 17.6% with 7 of 12 boards
- 44.2
- Index
- = 50.4
Score: the result on the board’s own 0 to 100 scale (dot; tick: the board’s leader). Board runs: how far the board’s scores sit above or below the same models’ scores on the other boards. Adjusted: the score minus that offset. The index is the adjusted average with the field average mixed in, more of it the fewer boards a model has. Each result links to its quote and page; the line under it is the setting it was measured with.
Sources
The quote, page and document for each of the 8 results
28HealthBench Professional Artificial Analysis… 0.532
Printed as 53.2%Independent run, measured Oct 2026Configuration: Artificial Analysis independent run, max reasoning; GPT-5.4 (low) judge, 4 repeats per task; length-adjusted (56.9 raw, 3,250 mean response chars)HealthBench Professional Benchmark | Artificial Analysis independent leaderboard, Artificial Analysis, 2 Nov 2026. Score section, Length-Adjusted Score chart, Kimi K3 (max) bar label; raw score 56.9% on the Raw Score tabHealthBench Professional: Length-Adjusted Score | Kimi K3 (max) | 53.2%
Every result from this document13MedScribe (Vals AI) 87.96
Printed as 87.96%Official leaderboard, measured Sep 2026Vals AI MedScribe leaderboard official leaderboard, Vals AI, 3 Sep 2026. Vals AI MedScribe leaderboard, View: All Models, Task: Overall, row 13 of 105 (Kimi K3), Accuracy column; Updated 9/29/2026. Rendered BenchmarkView table; configuration from embedded astro-island BenchmarkView props, benchmarkView.default.tasks.overall["kimi/kimi-k3"].13 | Kimi K3 | 87.96%±1.89 | $3/$15 | 2m17s
Every result from this document25MedCode (Vals AI) kimi-k3 48.88
Printed as 48.88%Official leaderboard, measured Sep 2026Configuration: model ID kimi/kimi-k3; temperature=1; max_output_tokens=30000Vals AI MedCode leaderboard official leaderboard, Vals AI, 29 Sep 2026. MedCode leaderboard, Overall task, All Models expanded, rank 24 of 103; Accuracy column; board Updated 9/29/2026. Rendered table row; configuration from embedded BenchmarkView props benchmarkView.default.tasks.overall["kimi/kimi-k3"].24 | Kimi K3 | 48.88%±2.19 | $3/$15 | 116.18s
Every result from this document2MAST (Medical AI Superintelligence Test) 60.1
Printed as 60.1%Official leaderboard, measured Aug 2026MAST: Medical AI Superintelligence Test leaderboard (General board) official leaderboard, ARISE AI Research Network, 15 Aug 2026. arise-ai.org/mast, 'Which AI can you trust for medical questions?' General tab, composite score table (8 of 11 models shown; 'Last updated August 15, 2026')2 Kimi K3 Moonshot AI 60.1%
Every result from this document7First, Do NOHARM (v2) 74.0
Printed as 74.0%Official leaderboard, measured Aug 2026MAST technical leaderboard (First Do NOHARM v2 and per-benchmark results) official leaderboard, ARISE AI Research Network, 15 Aug 2026. arise-ai.org/mast/technical, 'First Do NOHARM v2 overall metric across 19 models' ranking (Latest Flagships view), plus Model Leaderboard SAFETY column7Kimi K3OSSMoonshot AI 74.0%
Every result from this document7CHI-Bench 25.3
Printed as 25.3%Official leaderboard, measured Jul 2026CHI-Bench leaderboard (actAVA) official leaderboard, actAVA, 12 Aug 2026. CHI-Bench v1.0.0, All Domains, rank 08, Agent/Model and Accuracy columns; PA/UM/CM follow; board last updated 2026-08-12; board Date column 2026-07-24; accessed 2026-09-30. Linked submission provenance: run 2026-07-22T08:03:23.788878Z to 2026-07-22T14:22:38.269923Z; submitted_at 2026-08-12T20:20:00Z.08 | openai-agents | kimi-k3 | Open-source | 25.3% | 28.0% | 32.0% | 16.0% | 2026-07-24
Every result from this document- Also printed in CHI-Bench submission manifest: openai-agents + moonshotai/kimi-k3, submission.model; provenance.started_at / finished_at; results.overall.pass_at_1; retrieved 2026-09-30
6Bedside Bench Score 89.7
Printed as 89.7Official leaderboard, measured Oct 2026Configuration: Score (avg@3)Bedside Bench on the Fireworks Specialized Intelligence Index official leaderboard, Fireworks AI, 22 Oct 2026. https://fireworks.ai/specialized-intelligence-index/benchmarks/healthcare/bedside-bench (feed: benchmarks[id=healthcare/bedside-bench].results)Kimi K3 89.7 (Score (avg@3)), rank 7
Every result from this document- Also printed in Fireworks Specialized Intelligence Index (machine-readable feed): 89.7
4APEX-1 GP Score 64.35
Printed as 64.35Official leaderboard, measured Oct 2026Configuration: Score (avg@4), reasoning maxAPEX-1: General Practitioner on the Fireworks Specialized Intelligence Index official leaderboard, Fireworks AI, 26 Oct 2026. https://fireworks.ai/specialized-intelligence-index/benchmarks/healthcare/apex-1-general-practitioner (feed: benchmarks[id=healthcare/apex-1-general-practitioner].results)Kimi K3 64.35 (Score (avg@4)), rank 4
Every result from this document- Also printed in Fireworks Specialized Intelligence Index (machine-readable feed): 64.35
Other Moonshot AI models: Kimi K2.5, Kimi K2.6, Kimi K3.1 Pro