FIRE-Bench: leaderboard
Metric: Claim-level F1 (%; agent conclusions and the source paper's findings are split into atomic claims and matched by semantic entailment with a GPT-5.2 judge; mean over the 30 core LLM-behaviour rediscovery tasks and three runs; agent runs full research cycles from the research question in a GPU sandbox). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4 (Claude Code) | 46.7 |
| 2 | GPT-5 (Codex CLI, medium) | 41.9 |
| 3 | GPT-5 (OpenHands) | 37.9 |
| 4 | o4-mini (OpenHands) | 31.9 |
Interactive version: theaggregate.ai/benchmark?slug=fire-bench · How It Works · Data refreshed daily, snapshot 2026-09-26.