HealthBench Professional raw score (HealthBench Professional (raw))
We show this table for reference; we do not rank on it.
Raw score on physician-authored clinical consult, documentation, and research conversations.
Benchmark score on HealthBench Professional (raw) — November 13, 2026
We compile the HealthBench Professional (raw) rows from provider self-reports. Claude Sonnet 5.5 leads the table at 77.1%, followed by Claude Opus 5.5 (77.1%) and Claude Opus 5 (73.4%). We do not use these results to rank models overall.
Claude Sonnet 5.5
Anthropic
Reported by AnthropicClaude Opus 5.5
Anthropic
Reported by AnthropicClaude Opus 5
Anthropic
Reported by Anthropic8 modelsKnowledgeCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (8 models)
All 8 results on this page were reported by the labs that built the models.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5AnthropicReported by Anthropic | 77.1% | Reported by Anthropic | Not reported | Closed |
| 2 | Claude Opus 5.5AnthropicReported by Anthropic | 77.1% | Reported by Anthropic | Not reported | Closed |
| 3 | Claude Opus 5AnthropicReported by Anthropic | 73.4% | Reported by Anthropic | Not reported | Closed |
| 4 | GPT-6 AstraOpenAIReported by OpenAI | 68.2% | Reported by OpenAI | Not reported | Closed |
| 5 | GPT-6.1 SolOpenAIReported by OpenAI | 67.2% | Reported by OpenAI | Not reported | Closed |
| 6 | GPT-6.1 TerraOpenAIReported by OpenAI | 64.2% | Reported by OpenAI | Not reported | Closed |
| 7 | GPT-6 LunaOpenAIReported by OpenAI | 61.2% | Reported by OpenAI | Not reported | Closed |
| 8 | GPT-6 SolOpenAIReported by OpenAI | 59.5% | Reported by OpenAI | Not reported | Closed |
- Lab-reported
- Benchmark-run
- Independent
HealthBench Professional (raw, Artificial Analysis)
Artificial Analysis evaluates 525 conversations with GPT-5.4 (low) as judge and 4 repeats per task. Raw rubric score is reported in percentage points. Length adjustment uses 1.47 points per 500 answer characters above or below 2,000; reasoning is excluded. Effort and Default Fallback labels identify the evaluated configuration. These rows remain separate from provider runs and outside overall and category rankings.
| Source rank | Source model or configuration | Raw rubric score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 (Max, Default Fallback)Anthropic | 71.4% | Max, Default FallbackOriginal result | Not reported | Closed |
| 2 | Claude Fable 5.2 (Max, Default Fallback)Anthropic | 71.1% | Max, Default FallbackOriginal result | Not reported | Closed |
| 3 | Claude Sonnet 5.5 (Max, Default Fallback)Anthropic | 70.8% | Max, Default FallbackOriginal result | Not reported | Closed |
| 4 | GPT-6 Astra (Max)OpenAI | 69.7% | MaxOriginal result | Not reported | Closed |
| 5 | Claude Fable 5.1 (Max, Default Fallback)Anthropic | 69.2% | Max, Default FallbackOriginal result | Not reported | Closed |
| 6 | GPT-6.1 Sol (Max)OpenAI | 68.3% | MaxOriginal result | Not reported | Closed |
| 7 | Claude Haiku 5.5 (Max)Anthropic | 64.8% | MaxOriginal result | Not reported | Closed |
| 8 | Gemini 4 Argon (High)Google | 64.2% | HighOriginal result | Not reported | Closed |
| 9 | GPT-6 Luna (Max)OpenAI | 62.7% | MaxOriginal result | Not reported | Closed |
| 10 | GPT-6 Sol (Max)OpenAI | 61.2% | MaxOriginal result | Not reported | Closed |
| 11 | Gemini 3.8 Flash (High)Google | 59.0% | HighOriginal result | Not reported | Closed |
| 12 | GLM-5.3 (Max)Z.AI | 57.9% | MaxOriginal result | Not reported | Open |
| 13 | Grok 4.7 (Xhigh)xAI | 57.4% | XhighOriginal result | Not reported | Closed |
| 14 | Kimi K3 (Max)Moonshot AI | 56.9% | MaxOriginal result | Not reported | Open |
| 15 | DeepSeek V4.1 Flash (Max)DeepSeek | 52.9% | MaxOriginal result | Not reported | Open |
Among the reported HealthBench Professional (raw) rows, Claude Sonnet 5.5 is first at 77.1%. The third row is 3.7 points behind. The broader top-10 range is 17.6 points, so the table still separates the published systems.
8 models have been evaluated on HealthBench Professional (raw). The benchmark falls in the Knowledge category. HealthBench Professional (raw) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About HealthBench Professional (raw)
Year
2026
Tasks
525 physician-authored conversations
Format
Raw rubric score
Difficulty
Professional clinical tasks
This table retains unadjusted HealthBench Professional results from source-backed provider reports. Artificial Analysis independently evaluates the 525 conversations with four repeats per task and GPT-5.4 (low) as judge; its raw results appear below and stay separate from length-adjusted scores.
Freshness and provenance
Version
HealthBench Professional (raw) 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does HealthBench Professional (raw) measure?
Raw score on physician-authored clinical consult, documentation, and research conversations.
Which model scores highest on HealthBench Professional (raw)?
Claude Sonnet 5.5 by Anthropic currently leads with a score of 77.1% on HealthBench Professional (raw).
How many models are evaluated on HealthBench Professional (raw)?
8 AI models have published results on HealthBench Professional (raw) in the BenchLM catalog.
Compare top models on HealthBench Professional (raw)
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.