HealthBench Professional

Moonshot AI logoKimi K3 on HealthBench Professional

Rank 28 of 43, updated November 2, 2026.

Kimi K3 scores 0.532 on HealthBench Professional, rank 28 of 43 models on the board. HealthBench Professional scores models on 525 tasks drawn from real clinician conversations, graded criterion by criterion against physician-written rubrics on a 0 to 1 scale.

28
rank of 43
0.532
length-adjusted score
0.569
score before the length adjustment
0.171
gap to the leader, GPT-6 Astra (Anthropic run)
1M
context window, tokens
$3.00 / $15.00
input / output price per 1M tokens

Where it sits on the board

The models ranked around Kimi K3, by length-adjusted score. Differences are against Kimi K3.

RankModelLength-adjustedBefore the adjustmentDifference
25Google logoGemini 4 Argon0.5510.642+0.019
26Meta logoMuse Spark0.541–+0.009
27OpenAI logoGPT-5.6 Sol (August)0.5400.566+0.008
28Moonshot AI logoKimi K30.5320.569–
29Zhipu logoGLM 5.30.5280.579-0.004
30Anthropic logoClaude Opus 4.70.519–-0.013
31OpenAI logoGPT-5.50.5180.572-0.014

Result and API facts

Rank28 of 43
Length-adjusted score0.532
Score before the length adjustment0.569
LabMoonshot AI
Context window1M tokens
API price per 1M tokens$3.00 in / $15.00 out
LicenseKimi K3 License (open weights)
SourceHealthBench Professional Benchmark | Artificial Analysis (independent leaderboard)
Parameters2.8T total / 104B active
Released2026-07-16

Source of this score

Read from HealthBench Professional Benchmark | Artificial Analysis (independent leaderboard, Artificial Analysis, 2026-11-02). Independent run. Confidence: partial. Configuration: 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).

Score section, Length-Adjusted Score chart, Kimi K3 (max) bar label; raw score 56.9% on the Raw Score tabfull entry on the sources page
HealthBench Professional: Length-Adjusted Score | Kimi K3 (max) | 53.2%

Position in the field

The gap to the leader, GPT-6 Astra (Anthropic run) at 0.703, is 0.171. Directly above sits GPT-5.6 Sol (August) at 0.540. Directly below sits GLM 5.3 at 0.528. The scores on this page are compiled from published documents rather than from one controlled run, so a small gap between 2 models can reflect a difference in grader version, reasoning effort, or deployment setting as well as a difference in capability.

What does Kimi K3 score on HealthBench Professional?

Kimi K3 scores 0.532 on HealthBench Professional, which places it at rank 28 of 43 models on the board as of November 2, 2026. The number was read from HealthBench Professional Benchmark | Artificial Analysis, listed on the sources page.

How much does Kimi K3 cost to run?

Kimi K3 is priced at $3.00 per million input tokens and $15.00 per million output tokens through Moonshot AI's API.

Pairings with a dedicated comparison page are linked; every other difference is in the score-difference matrix.

  • Kimi K3 vs GPT-6 Astra (Anthropic run)
    0.532 vs 0.703, GPT-6 Astra (Anthropic run) by 0.171
  • Kimi K3 vs Claude Sonnet 5.5
    0.532 vs 0.692, Claude Sonnet 5.5 by 0.160
  • Kimi K3 vs Claude Fable 5
    0.532 vs 0.660, Claude Fable 5 by 0.128
  • Kimi K3 vs Claude Opus 5.5
    0.532 vs 0.656, Claude Opus 5.5 by 0.124
  • Kimi K3 vs GPT-6 Luna (Anthropic run)
    0.532 vs 0.655, GPT-6 Luna (Anthropic run) by 0.123
  • Kimi K3 vs Claude Haiku 5.5
    0.532 vs 0.648, Claude Haiku 5.5 by 0.116
  • Kimi K3 vs GPT-6 Astra
    0.532 vs 0.647, GPT-6 Astra by 0.115
  • Kimi K3 vs GPT-6.1 Sol
    0.532 vs 0.642, GPT-6.1 Sol by 0.110
  • Kimi K3 vs Claude Fable 5 (September card)
    0.532 vs 0.633, Claude Fable 5 (September card) by 0.101
  • Kimi K3 vs GPT-6.1 Terra
    0.532 vs 0.627, GPT-6.1 Terra by 0.095
  • Kimi K3 vs Muse Wonder Full
    0.532 vs 0.623, Muse Wonder Full by 0.091
  • Kimi K3 vs Claude Fable 5.1
    0.532 vs 0.621, Claude Fable 5.1 by 0.089
  • Kimi K3 vs GPT-6 Luna
    0.532 vs 0.608, GPT-6 Luna by 0.076
  • Kimi K3 vs GPT-6 Sol
    0.532 vs 0.608, GPT-6 Sol by 0.076
  • Kimi K3 vs GPT-5.6 Sol
    0.532 vs 0.605, GPT-5.6 Sol by 0.073
  • Kimi K3 vs Muse Wonder
    0.532 vs 0.605, Muse Wonder by 0.073
  • Kimi K3 vs Claude Opus 5
    0.532 vs 0.598, Claude Opus 5 by 0.066
  • Kimi K3 vs Muse Spark 1.1
    0.532 vs 0.593, Muse Spark 1.1 by 0.061
  • Kimi K3 vs Claude Sonnet 5
    0.532 vs 0.578, Claude Sonnet 5 by 0.046
  • Kimi K3 vs GPT-5.6 Terra
    0.532 vs 0.577, GPT-5.6 Terra by 0.045
  • Kimi K3 vs Claude Opus 4.8 (Opus 4.8 grader)
    0.532 vs 0.574, Claude Opus 4.8 (Opus 4.8 grader) by 0.042
  • Kimi K3 vs Grok 4.7
    0.532 vs 0.567, Grok 4.7 by 0.035
  • Kimi K3 vs Claude Opus 4.8
    0.532 vs 0.558, Claude Opus 4.8 by 0.026
  • Kimi K3 vs GPT-5.6 Luna
    0.532 vs 0.557, GPT-5.6 Luna by 0.025
  • Kimi K3 vs Gemini 4 Argon
    0.532 vs 0.551, Gemini 4 Argon by 0.019
  • Kimi K3 vs Muse Spark
    0.532 vs 0.541, Muse Spark by 0.009
  • Kimi K3 vs GPT-5.6 Sol (August)
    0.532 vs 0.540, GPT-5.6 Sol (August) by 0.008
  • Kimi K3 vs GLM 5.3
    0.532 vs 0.528, Kimi K3 by 0.004
  • Kimi K3 vs Claude Opus 4.7
    0.532 vs 0.519, Kimi K3 by 0.013
  • Kimi K3 vs GPT-5.5
    0.532 vs 0.518, Kimi K3 by 0.014
  • Kimi K3 vs Gemini 3.8 Flash
    0.532 vs 0.517, Kimi K3 by 0.015
  • Kimi K3 vs DeepSeek V4.1 Flash
    0.532 vs 0.507, Kimi K3 by 0.025
  • Kimi K3 vs Grok 4.6
    0.532 vs 0.485, Kimi K3 by 0.047
  • Kimi K3 vs GPT-5.4
    0.532 vs 0.481, Kimi K3 by 0.051
  • Kimi K3 vs GPT-5
    0.532 vs 0.462, Kimi K3 by 0.070
  • Kimi K3 vs GPT-5.2
    0.532 vs 0.459, Kimi K3 by 0.073
  • Kimi K3 vs Claude Sonnet 4.6
    0.532 vs 0.442, Kimi K3 by 0.090
  • Kimi K3 vs GPT-5.6 Luna (August)
    0.532 vs 0.441, Kimi K3 by 0.091
  • Kimi K3 vs GPT-5.1
    0.532 vs 0.396, Kimi K3 by 0.136
  • Kimi K3 vs GPT-5.5 Instant
    0.532 vs 0.384, Kimi K3 by 0.148
  • Kimi K3 vs MAI-Thinking-1
    0.532 vs 0.350, Kimi K3 by 0.182
  • Kimi K3 vs Claude Haiku 4.5
    0.532 vs 0.322, Kimi K3 by 0.210

Where the scores come from and how they are read is on the methodology page. The full ranking is on the leaderboard.