LABBench2Browse 284 families
LABBench2
Can the model reason accurately about difficult scientific material?
Version / measure
- Domain
- Science
- Ranked models
- 7
- Reported values
- 10
- Reports
- 2 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Gemini 3.8 FlashGoogle | reportedcost n/a1 value · Sep 2 · source | 86.2% |
| 2 | Claude Opus 5Anthropic | reportedcost n/a1 value · Sep 2 · source | 84.2% |
| 3 | Gemini 3.7 FlashGoogle | reportedcost n/a2 values · Sep 2 · source | 82.1% |
| 4 | GPT-5.6 SolOpenAI | reportedcost n/a1 value · Sep 2 · source | 82.1% |
| 5 | GPT-5.6 TerraOpenAI | reportedcost n/a2 values · Sep 2 · source | 81.2% |
| 6 | Claude Sonnet 5Anthropic | reportedcost n/a2 values · Sep 2 · source | 80.1% |
| 7 | Gemini 3.6 FlashGoogle | reportedcost n/a1 value · Aug 13 · source | 76.1% |
Effort curve
Score vs. reported cost. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology
Can the model reason accurately about difficult scientific material?
Scientific knowledge and reasoning under the benchmark's published question set. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
Google-computed for Gemini and GPT-5.6 Terra with a Linux terminal, bioinformatics tools, Python, R, and internet access. Biology real-world research tasks; exact model-card cells.
Method / source ↗