GPT-6 Astra (Anthropic run) on HealthBench Professional
rank 1 of 31 · updated September 30, 2026
GPT-6 Astra (Anthropic run) scores 0.703 on HealthBench Professional, rank 1 of 31 models on the board. The first GPT-6 model, shipped September 3, 2026, at twice the input price of GPT-5.6 Sol. 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 | 1 of 31 |
|---|---|
| score | 0.703 |
| lab | OpenAI |
| context window | 1.05M tokens |
| API price per 1M tokens | $10.00 in / $50.00 out |
| license | proprietary |
| source | Claude Sonnet 5.5 System Card (system card) |
| released | 2026-09-03 |
Position in the field
GPT-6 Astra (Anthropic run) leads the leaderboard, 0.011 ahead of Claude Sonnet 5.5 in second place. 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 Claude Sonnet 5.5 System Card (system card, Anthropic, 2026-09-28). Independent run. Confidence: verified. Configuration: Anthropic public-API reproduction; max effort; no system prompt; Opus 4.8 grader; length-adjusted; raw 74.0%.
On HealthBench Professional at max effort, length adjustment changes the ranking. After: GPT-6 Astra (70.3%) > Claude Sonnet 5.5 (69.2%) > Claude Opus 5.5 (65.6%) > Claude Fable 5.1 (62.1%).
p. 138, paragraph above Figure 8.15.B, HealthBench Professional after length adjustment. · full entry on the sources page
What does GPT-6 Astra (Anthropic run) score on HealthBench Professional?
GPT-6 Astra (Anthropic run) scores 0.703 on HealthBench Professional, which places it at rank 1 of 31 models on the board as of September 30, 2026. The number was read from Claude Sonnet 5.5 System Card, listed on the sources page.
How much does GPT-6 Astra (Anthropic run) cost to run?
GPT-6 Astra (Anthropic run) is priced at $10.00 per million input tokens and $50.00 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-6 Astra (Anthropic run) vs Claude Sonnet 5.50.703 vs 0.692 · GPT-6 Astra (Anthropic run) by 0.011
- GPT-6 Astra (Anthropic run) vs Claude Fable 50.703 vs 0.660 · GPT-6 Astra (Anthropic run) by 0.043
- GPT-6 Astra (Anthropic run) vs Claude Opus 5.50.703 vs 0.656 · GPT-6 Astra (Anthropic run) by 0.047
- GPT-6 Astra (Anthropic run) vs GPT-6 Astra0.703 vs 0.647 · GPT-6 Astra (Anthropic run) by 0.056
- GPT-6 Astra (Anthropic run) vs Claude Fable 5 (September card)0.703 vs 0.633 · GPT-6 Astra (Anthropic run) by 0.070
- GPT-6 Astra (Anthropic run) vs Claude Fable 5.10.703 vs 0.621 · GPT-6 Astra (Anthropic run) by 0.082
- GPT-6 Astra (Anthropic run) vs GPT-6 Luna0.703 vs 0.608 · GPT-6 Astra (Anthropic run) by 0.095
- GPT-6 Astra (Anthropic run) vs GPT-6 Sol0.703 vs 0.608 · GPT-6 Astra (Anthropic run) by 0.095
- GPT-6 Astra (Anthropic run) vs GPT-5.6 Sol0.703 vs 0.605 · GPT-6 Astra (Anthropic run) by 0.098
- GPT-6 Astra (Anthropic run) vs Claude Opus 50.703 vs 0.598 · GPT-6 Astra (Anthropic run) by 0.105
- GPT-6 Astra (Anthropic run) vs Muse Spark 1.10.703 vs 0.593 · GPT-6 Astra (Anthropic run) by 0.110
- GPT-6 Astra (Anthropic run) vs Claude Sonnet 50.703 vs 0.578 · GPT-6 Astra (Anthropic run) by 0.125
- GPT-6 Astra (Anthropic run) vs GPT-5.6 Terra0.703 vs 0.577 · GPT-6 Astra (Anthropic run) by 0.126
- GPT-6 Astra (Anthropic run) vs Claude Opus 4.8 (Opus 4.8 grader)0.703 vs 0.574 · GPT-6 Astra (Anthropic run) by 0.129
- GPT-6 Astra (Anthropic run) vs Grok 4.70.703 vs 0.567 · GPT-6 Astra (Anthropic run) by 0.136
- GPT-6 Astra (Anthropic run) vs Claude Opus 4.80.703 vs 0.558 · GPT-6 Astra (Anthropic run) by 0.145
- GPT-6 Astra (Anthropic run) vs GPT-5.6 Luna0.703 vs 0.557 · GPT-6 Astra (Anthropic run) by 0.146
- GPT-6 Astra (Anthropic run) vs Muse Spark0.703 vs 0.541 · GPT-6 Astra (Anthropic run) by 0.162
- GPT-6 Astra (Anthropic run) vs GPT-5.6 Sol (August)0.703 vs 0.540 · GPT-6 Astra (Anthropic run) by 0.163
- GPT-6 Astra (Anthropic run) vs Claude Opus 4.70.703 vs 0.519 · GPT-6 Astra (Anthropic run) by 0.184
- GPT-6 Astra (Anthropic run) vs GPT-5.50.703 vs 0.518 · GPT-6 Astra (Anthropic run) by 0.185
- GPT-6 Astra (Anthropic run) vs Grok 4.60.703 vs 0.485 · GPT-6 Astra (Anthropic run) by 0.218
- GPT-6 Astra (Anthropic run) vs GPT-5.40.703 vs 0.481 · GPT-6 Astra (Anthropic run) by 0.222
- GPT-6 Astra (Anthropic run) vs GPT-50.703 vs 0.462 · GPT-6 Astra (Anthropic run) by 0.241
- GPT-6 Astra (Anthropic run) vs GPT-5.20.703 vs 0.459 · GPT-6 Astra (Anthropic run) by 0.244
- GPT-6 Astra (Anthropic run) vs Claude Sonnet 4.60.703 vs 0.442 · GPT-6 Astra (Anthropic run) by 0.261
- GPT-6 Astra (Anthropic run) vs GPT-5.6 Luna (August)0.703 vs 0.441 · GPT-6 Astra (Anthropic run) by 0.262
- GPT-6 Astra (Anthropic run) vs GPT-5.10.703 vs 0.396 · GPT-6 Astra (Anthropic run) by 0.307
- GPT-6 Astra (Anthropic run) vs GPT-5.5 Instant0.703 vs 0.384 · GPT-6 Astra (Anthropic run) by 0.319
- GPT-6 Astra (Anthropic run) vs MAI-Thinking-10.703 vs 0.350 · GPT-6 Astra (Anthropic run) by 0.353
Where the scores come from and how they are read is on the methodology page. The full ranking is on the leaderboard.