<!-- Generated by scripts/build-backrooms-results-table.mjs. Do not edit directly; see docs/updating-results.md. -->

# RareDiseaseBench results

> RareDiseaseBench tests whether AI agents can integrate clinical, multiomic, and functional data to make decisions around diagnosing and managing rare disease. In each evaluation, agents receive a clinical presentation and must request information from an isolated filesystem, testing their ability not only to analyze data but to brainstorm what is required. Rubric-graded scores assess how often agents reach the right diagnosis and management plan. Refusals, errors and unfinished runs are failures.

Data updated: 2026-11-05

Artifact generated: 2026-11-06T17:31:23.387Z

- Domain: Applications
- Horizon: long
- Assessment: Capabilities
- Published score scope: Current V1 evaluation set (35 evaluations)

Scores use a 0–100 percentage scale. Consult the repository and paper for benchmark-specific task construction, grading, aggregation, and uncertainty methodology. Results compare model-and-harness combinations.

LSVP: Anthropic's safety-filter-exempt evaluation variant, which produces fewer refusals on biology prompts. LSVP results are reported alongside the model's standard run and are not a replacement for it.

| Rank | Model | Harness | Provider | Score | Variant |
| ---: | --- | --- | --- | ---: | --- |
| 1 | Gemini 4 Argon | Pi | Google | 43.8% |  |
| 2 | Claude Opus 5.5* | Claude Code | Anthropic | 33.3% | LSVP |
| 3 | GPT-6 Astra | Pi | OpenAI | 33.3% |  |
| 4 | Claude Opus 5.5* | Pi | Anthropic | 28.6% | LSVP |
| 5 | GPT-6.1 Sol | Pi | OpenAI | 27.6% |  |
| 6 | GPT-6 Astra | Codex | OpenAI | 25.7% |  |
| 7 | GPT-6.1 Sol | Codex | OpenAI | 25.7% |  |
| 8 | GPT-6.1 Terra | Pi | OpenAI | 25.7% |  |
| 9 | GPT-6.1 Terra | Codex | OpenAI | 23.8% |  |
| 10 | Claude Sonnet 5.5* | Claude Code | Anthropic | 22.9% | LSVP |
| 11 | Claude Sonnet 5.5* | Pi | Anthropic | 20.9% | LSVP |
| 12 | Claude Fable 5.1* | Claude Code | Anthropic | 19.1% | LSVP |
| 13 | Claude Fable 5.1* | Pi | Anthropic | 13.3% | LSVP |
| 14 | GPT-5.6 Sol | Pi | OpenAI | 13.3% |  |
| 15 | Claude Opus 5* | Claude Code | Anthropic | 11.4% | LSVP |
| 16 | Claude Opus 5* | Pi | Anthropic | 11.4% | LSVP |
| 17 | GPT-5.6 Sol | Codex | OpenAI | 4.8% |  |
| 18 | Grok 4.6 | Pi | SpaceXAI | 4.8% |  |
| 19 | Gemini 3.8 Flash | Pi | Google | 2.9% |  |
| 20 | Mistral Large 4.0 | Pi | Mistral | 1.9% |  |
| 21 | Claude Sonnet 5* | Claude Code | Anthropic | 0.9% | LSVP |
| 22 | GPT-5.6 Luna | Pi | OpenAI | 0.9% |  |
| 23 | Claude Sonnet 5* | Pi | Anthropic | 0.0% | LSVP |
| 24 | DeepSeek V4.1 Flash | Pi | DeepSeek | 0.0% |  |
| 25 | GLM 5.3 | Pi | Z.ai | 0.0% |  |
| 26 | GPT-5.6 Luna | Codex | OpenAI | 0.0% |  |
| 27 | GPT-6 Luna | Codex | OpenAI | 0.0% |  |
| 28 | GPT-6 Luna | Pi | OpenAI | 0.0% |  |
| 29 | Kimi K3 | Pi | Moonshot | 0.0% |  |
| 30 | Nemotron 3 Ultra 550B A55B | Pi | NVIDIA | 0.0% |  |
| 31 | Nemotron 3.5 Lightning | Pi | NVIDIA | 0.0% |  |

## Sources and downloads

- [Paper PDF](/papers/rarediseasebench.pdf)
- [Original publication](https://latch.bio/rarediseasebench)
- [Repository](https://github.com/latchbio/rarediseasebench)
- [Interactive benchmark](https://benchmarks.bio/raredisease)
- [JSON](https://benchmarks.bio/benchmarks/rarediseasebench/latest.json)
- [CSV](https://benchmarks.bio/benchmarks/rarediseasebench/latest.csv)
- [Current V1 cross-benchmark leaderboard](https://benchmarks.bio/results/v1/)