Compare current model-and-harness results for TissueMicroscopyBench. TissueMicroscopyBench tests whether AI agents can analyze real tissue microscopy images. Tasks include locating cells and structures, quantifying staining and morphology, and tracking changes over time and after treatment. Each pairs real image data with a specific analysis question and a verifiable answer. Refusals, errors and unfinished runs are failures. Current V1 covers 49 evaluations and 30 ranked configurations. Data updated 2026-11-10. The highest-ranked published configurations are Claude Fable 5.2* (Pi, LSVP, 34.0%); Claude Fable 5.2* (Claude Code, LSVP, 32.6%); Claude Opus 5.5* (Pi, LSVP, 29.3%). The highest-ranked published run without an LSVP exemption is GPT-6.1 Terra (Pi, 25.9%). Scores report benchmark-specific biological task performance on a 0–100 percentage scale. See the linked paper and repository for task construction, grading and uncertainty methodology. Results describe the complete model and agent harness used for each run, and are not comparable across different benchmarks or evaluation versions. 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 | Claude Fable 5.2* | Pi | Anthropic | 34.0% | LSVP |
| 2 | Claude Fable 5.2* | Claude Code | Anthropic | 32.6% | LSVP |
| 3 | Claude Opus 5.5* | Pi | Anthropic | 29.3% | LSVP |
| 4 | Claude Sonnet 5.5* | Pi | Anthropic | 27.9% | LSVP |
| 5 | Claude Opus 5.5* | Claude Code | Anthropic | 25.9% | LSVP |
| 6 | GPT-6.1 Terra | Pi | OpenAI | 25.9% | |
| 7 | Claude Fable 5.1* | Pi | Anthropic | 24.5% | LSVP |
| 8 | Claude Sonnet 5.5* | Claude Code | Anthropic | 24.5% | LSVP |
| 9 | GPT-6.1 Sol | Pi | OpenAI | 22.4% | |
| 10 | Gemini 4 Argon | Pi | 21.8% | ||
| 11 | GPT-6 Astra | Pi | OpenAI | 19.7% | |
| 12 | Claude Fable 5.1* | Claude Code | Anthropic | 17.0% | LSVP |
| 13 | Claude Opus 5* | Claude Code | Anthropic | 15.7% | LSVP |
| 14 | Claude Opus 5* | Pi | Anthropic | 14.3% | LSVP |
| 15 | Grok 4.7 | Pi | SpaceXAI | 6.8% | |
| 16 | Grok 4.7 | Grok Build | SpaceXAI | 5.4% | |
| 17 | Grok 4.6 | Grok Build | SpaceXAI | 4.1% | |
| 18 | Gemini 3.8 Flash | Pi | 3.4% | ||
| 19 | DeepSeek V4.1 Flash | Pi | DeepSeek | 2.7% | |
| 20 | Grok 4.6 | Pi | SpaceXAI | 2.7% | |
| 21 | Claude Sonnet 5* | Claude Code | Anthropic | 2.0% | LSVP |
| 22 | GPT-5.6 Sol | Pi | OpenAI | 2.0% | |
| 23 | Mistral Large 4.0 | Pi | Mistral | 2.0% | |
| 24 | Claude Sonnet 5* | Pi | Anthropic | 1.4% | LSVP |
| 25 | GLM 5.3 | Pi | Z.ai | 1.4% | |
| 26 | GPT-6 Luna | Pi | OpenAI | 1.4% | |
| 27 | GPT-5.6 Luna | Pi | OpenAI | 0.7% | |
| 28 | Kimi K3 | Pi | Moonshot | 0.7% | |
| 29 | Nemotron 3 Ultra 550B A55B | Pi | NVIDIA | 0.0% | |
| 30 | Nemotron 3.5 Lightning | Pi | NVIDIA | 0.0% |
LSVP access
Anthropic's Life Sciences Verification Program accepts interest from verified life sciences organizations using Claude Team, Enterprise or the first-party API. Eligibility, supported models and grant types are determined by Anthropic. Program access does not establish equivalence to every benchmark evaluation variant. Guidance checked 2026-11-10. Anthropic eligibility guidance · Register organizational interest.
Methodology and provenance
Scores use a 0–100 percentage scale. Publication links are not yet available.
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.