GPT-6 Astra
- Made by
- OpenAI
- Released
- 3 September 2026
- Licence
- proprietary
- Context
- 1.05M
- Input, per 1M tokens
- $10
- Output, per 1M tokens
- $50
How this score is made
- HealthBench Professional0.70370.3 out of 100runs −2.5 hardadjusted 72.8weight 0.446, supportingsystem cardAnthropic public-API reproduction; max effort; no system prompt; Opus 4.8 grader; length-adjusted; raw 74.0%
- HealthBench58.358.3 out of 100runs −0.5 leveladjusted 58.8weight 0.798, coresystem cardReported by the maker.length-adjusted, max reasoning effort (59.7 unadjusted, 2,258 mean response chars); GPT-6 Astra system card Table 6, column 'gpt-6 Astra'.
- Health Optimization Bench: incretin82.282.2 out of 100runs +4.0 easyadjusted 78.2weight 0.226, supportingofficial leaderboardNear its ceiling (leader 84.9), so it counts for less.max reasoning effort
- SAFE-TeenNo result here, and none is filled in.
- Adjusted average of 3 benchmarks, by weight
- 66.0
- Field average, counts for 33.3% with 3 of 4 benchmarks
- 55.3
- Index
- = 62.4
Benchmark by benchmark
HealthBench Professional1 of 37
- 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
- 0.621Claude Fable 5.1
37 modelsFull benchmark
HealthBench14 of 26
- 68.1Claude Opus 5.5
- 67.1Claude Opus 5
- 65.4Claude Sonnet 5.5
- 65.1Baichuan-M3
- 63.3GPT-5.2-High
- 62.7Claude Fable 5
- 61.6Claude Haiku 5.5
- 60Claude Fable 5.1
- 59.3Claude Opus 4.8
- 58.3GPT-6 Astra
26 modelsFull benchmark
Health Optimization Bench: incretin evidence synthesis3 of 17
- 84.9Claude Fable 5.1
- 83.8Claude Fable 5
- 82.2GPT-6 Astra (max)
- 81.3Grok 4.6
- 78.3Claude Opus 5
- 77.9GPT-5.6 Sol (max)
- 77.8Kimi K3
- 68.4Muse Spark
- 67.8Gemini 3.8 Flash
- 57.9Gemini 3.6
17 modelsFull benchmark
Artificial Analysis Healthcare & Medical Index4 of 29
- 61Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)
- 58Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)
- 53Claude Opus 5 (Adaptive Reasoning, Max Effort)
- 52Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)
- 52GPT-6 Astra (max)
- 50Muse Spark 1.3 (max)
- 47GLM-5.3 (max)
- 47Grok 4.7 (xhigh)
- 45GLM 5.3 Flash
- 45GPT-5.6 Sol (max)
29 modelsFull benchmark