GPT-5.6 Sol (August) on HealthBench Professional

rank 12 of 22 · updated September 8, 2026

GPT-5.6 Sol (August) scores 0.540 on HealthBench Professional, rank 12 of 22 models on the board. The flagship tier of OpenAI's GPT-5.6 family, aimed at its hardest reasoning and agentic workloads. 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	12 of 22
score	0.540
lab	OpenAI
context window	1.1M tokens
API price per 1M tokens	$5.00 in / $30.00 out
license	proprietary
source	GPT-5.6 - August Updates (system card addendum) (system card)
released	2026-07-09

Position in the field

The gap to the leader, Claude Fable 5 at 0.660, is 0.120. Directly above sits Muse Spark at 0.541. Directly below sits Claude Opus 4.7 at 0.519. 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 GPT-5.6 - August Updates (system card addendum) (system card, OpenAI, 2026-08-06). Vendor-reported. Confidence: verified. Configuration: ChatGPT production (Instant) deployment setting, length-adjusted, GPT-5.6 August Updates PDF column GPT-5.6 Sol (August) (56.6 unadjusted, 2,894 chars).

HealthBench Professional 32.9 (33.8, 2,285) 38.4 (40.7, 2,775) 54.0 (56.6, 2,894) 44.1 (46.8, 2,920)

p. 11, section 5.1 HealthBench, table "Reported as length-adjusted score (unadjusted, mean response length in characters)", column GPT-5.6 Sol (August) · full entry on the sources page

What does GPT-5.6 Sol (August) score on HealthBench Professional?

GPT-5.6 Sol (August) scores 0.540 on HealthBench Professional, which places it at rank 12 of 22 models on the board as of September 8, 2026. The number was read from GPT-5.6 - August Updates (system card addendum), listed on the sources page.

How much does GPT-5.6 Sol (August) cost to run?

GPT-5.6 Sol (August) is priced at $5.00 per million input tokens and $30.00 per million output tokens through OpenAI's API.

Head to head

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

GPT-5.6 Sol (August) vs Claude Fable 5
0.540 vs 0.660 · Claude Fable 5 by 0.120
GPT-5.6 Sol (August) vs GPT-6 Astra
0.540 vs 0.634 · GPT-6 Astra by 0.094
GPT-5.6 Sol (August) vs Claude Fable 5.1
0.540 vs 0.621 · Claude Fable 5.1 by 0.081
GPT-5.6 Sol (August) vs GPT-5.6 Sol
0.540 vs 0.605 · GPT-5.6 Sol by 0.065
GPT-5.6 Sol (August) vs Claude Opus 5
0.540 vs 0.598 · Claude Opus 5 by 0.058
GPT-5.6 Sol (August) vs Muse Spark 1.1
0.540 vs 0.593 · Muse Spark 1.1 by 0.053
GPT-5.6 Sol (August) vs Claude Sonnet 5
0.540 vs 0.578 · Claude Sonnet 5 by 0.038
GPT-5.6 Sol (August) vs GPT-5.6 Terra
0.540 vs 0.577 · GPT-5.6 Terra by 0.037
GPT-5.6 Sol (August) vs Claude Opus 4.8
0.540 vs 0.558 · Claude Opus 4.8 by 0.018
GPT-5.6 Sol (August) vs GPT-5.6 Luna
0.540 vs 0.557 · GPT-5.6 Luna by 0.017
GPT-5.6 Sol (August) vs Muse Spark
0.540 vs 0.541 · Muse Spark by 0.001
GPT-5.6 Sol (August) vs Claude Opus 4.7
0.540 vs 0.519 · GPT-5.6 Sol (August) by 0.021
GPT-5.6 Sol (August) vs GPT-5.5
0.540 vs 0.518 · GPT-5.6 Sol (August) by 0.022
GPT-5.6 Sol (August) vs GPT-5.4
0.540 vs 0.481 · GPT-5.6 Sol (August) by 0.059
GPT-5.6 Sol (August) vs GPT-5
0.540 vs 0.462 · GPT-5.6 Sol (August) by 0.078
GPT-5.6 Sol (August) vs GPT-5.2
0.540 vs 0.459 · GPT-5.6 Sol (August) by 0.081
GPT-5.6 Sol (August) vs Claude Sonnet 4.6
0.540 vs 0.442 · GPT-5.6 Sol (August) by 0.098
GPT-5.6 Sol (August) vs GPT-5.6 Luna (August)
0.540 vs 0.441 · GPT-5.6 Sol (August) by 0.099
GPT-5.6 Sol (August) vs GPT-5.1
0.540 vs 0.396 · GPT-5.6 Sol (August) by 0.144
GPT-5.6 Sol (August) vs GPT-5.5 Instant
0.540 vs 0.384 · GPT-5.6 Sol (August) by 0.156
GPT-5.6 Sol (August) vs MAI-Thinking-1
0.540 vs 0.350 · GPT-5.6 Sol (August) by 0.190

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