Benchmarks

Overview

We evaluate AI agents on the following scientific benchmarks:

Benchmark	Description
GPQA	Graduate-level physics, chemistry, and biology questions (GPQA Diamond)
FrontierScience	Physics, chemistry, biology problems at Olympiad and research level
LAB-Bench	Biology research capabilities across 8 subcategories (LitQA, DbQA, FigQA, SuppQA, TableQA, ProtocolQA, SeqQA, Cloning Scenarios)
BixBench	Real-world biological data analysis scenarios

All results below are loaded dynamically from evaluation data.

---

Benchmark Statistics

Score distribution across models for each benchmark.

---

Benchmark Difficulty

Benchmarks sorted by average model score — lower scores indicate harder benchmarks.

---

Individual Benchmark Details

Select a benchmark to view detailed results:

-- Select a Benchmark --

---

Back to Leaderboard | View Methodology