Grok 4.6 on HealthBench Professional

rank 23 of 31 · updated September 30, 2026

Grok 4.6 scores 0.485 on HealthBench Professional, rank 23 of 31 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.

Result and API facts

rank	23 of 31
score	0.485
lab	SpaceX AI
context window	500K tokens
API price per 1M tokens	$2.00 in / $6.00 out
license	proprietary
source	Introducing Grok 4.7 (launch post)
released	2026-08-12

Position in the field

The gap to the leader, GPT-6 Astra (Anthropic run) at 0.703, is 0.218. Directly above sits GPT-5.5 at 0.518. Directly below sits GPT-5.4 at 0.481. 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.

Source of this score

Read from Introducing Grok 4.7 (launch post, SpaceXAI, 2026-09-21). Vendor-reported. Confidence: verified. Configuration: SpaceXAI evaluation; High effort; HealthBench Professional.

Grok 4.7 xHigh Grok 4.6 High GPT-5.6 Sol Max Fable 5.1 Max Clinical reasoningHealthBench Professional 56.7% 48.5% 60.5% 62.1%

Model Improvements comparison table, Clinical reasoning / HealthBench Professional row; Grok 4.6 column; September 21, 2026. · full entry on the sources page

What does Grok 4.6 score on HealthBench Professional?

Grok 4.6 scores 0.485 on HealthBench Professional, which places it at rank 23 of 31 models on the board as of September 30, 2026. The number was read from Introducing Grok 4.7, listed on the sources page.

How much does Grok 4.6 cost to run?

Grok 4.6 is priced at $2.00 per million input tokens and $6.00 per million output tokens through SpaceX AI's API.

Head to head

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

Grok 4.6 vs GPT-6 Astra (Anthropic run)
0.485 vs 0.703 · GPT-6 Astra (Anthropic run) by 0.218
Grok 4.6 vs Claude Sonnet 5.5
0.485 vs 0.692 · Claude Sonnet 5.5 by 0.207
Grok 4.6 vs Claude Fable 5
0.485 vs 0.660 · Claude Fable 5 by 0.175
Grok 4.6 vs Claude Opus 5.5
0.485 vs 0.656 · Claude Opus 5.5 by 0.171
Grok 4.6 vs GPT-6 Astra
0.485 vs 0.647 · GPT-6 Astra by 0.162
Grok 4.6 vs Claude Fable 5 (September card)
0.485 vs 0.633 · Claude Fable 5 (September card) by 0.148
Grok 4.6 vs Claude Fable 5.1
0.485 vs 0.621 · Claude Fable 5.1 by 0.136
Grok 4.6 vs GPT-6 Luna
0.485 vs 0.608 · GPT-6 Luna by 0.123
Grok 4.6 vs GPT-6 Sol
0.485 vs 0.608 · GPT-6 Sol by 0.123
Grok 4.6 vs GPT-5.6 Sol
0.485 vs 0.605 · GPT-5.6 Sol by 0.120
Grok 4.6 vs Claude Opus 5
0.485 vs 0.598 · Claude Opus 5 by 0.113
Grok 4.6 vs Muse Spark 1.1
0.485 vs 0.593 · Muse Spark 1.1 by 0.108
Grok 4.6 vs Claude Sonnet 5
0.485 vs 0.578 · Claude Sonnet 5 by 0.093
Grok 4.6 vs GPT-5.6 Terra
0.485 vs 0.577 · GPT-5.6 Terra by 0.092
Grok 4.6 vs Claude Opus 4.8 (Opus 4.8 grader)
0.485 vs 0.574 · Claude Opus 4.8 (Opus 4.8 grader) by 0.089
Grok 4.6 vs Grok 4.7
0.485 vs 0.567 · Grok 4.7 by 0.082
Grok 4.6 vs Claude Opus 4.8
0.485 vs 0.558 · Claude Opus 4.8 by 0.073
Grok 4.6 vs GPT-5.6 Luna
0.485 vs 0.557 · GPT-5.6 Luna by 0.072
Grok 4.6 vs Muse Spark
0.485 vs 0.541 · Muse Spark by 0.056
Grok 4.6 vs GPT-5.6 Sol (August)
0.485 vs 0.540 · GPT-5.6 Sol (August) by 0.055
Grok 4.6 vs Claude Opus 4.7
0.485 vs 0.519 · Claude Opus 4.7 by 0.034
Grok 4.6 vs GPT-5.5
0.485 vs 0.518 · GPT-5.5 by 0.033
Grok 4.6 vs GPT-5.4
0.485 vs 0.481 · Grok 4.6 by 0.004
Grok 4.6 vs GPT-5
0.485 vs 0.462 · Grok 4.6 by 0.023
Grok 4.6 vs GPT-5.2
0.485 vs 0.459 · Grok 4.6 by 0.026
Grok 4.6 vs Claude Sonnet 4.6
0.485 vs 0.442 · Grok 4.6 by 0.043
Grok 4.6 vs GPT-5.6 Luna (August)
0.485 vs 0.441 · Grok 4.6 by 0.044
Grok 4.6 vs GPT-5.1
0.485 vs 0.396 · Grok 4.6 by 0.089
Grok 4.6 vs GPT-5.5 Instant
0.485 vs 0.384 · Grok 4.6 by 0.101
Grok 4.6 vs MAI-Thinking-1
0.485 vs 0.350 · Grok 4.6 by 0.135

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