An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper.
BenchLM mirrors the official ResearchClawBench Pass@1 leaderboard snapshot. The source benchmark contains 40 tasks across 10 scientific domains and uses RADS average (Pass@1) as the primary metric.
ResearchClawBench gives agents related literature and raw data, hides the target paper, and grades how much of the scientific result they rediscover. The RADS scale treats 50 as matching the original paper and 70+ as surpassing it.
ResearchClawBench is display only on BenchLM. The rows combine a model, a research harness, execution budget, and long-horizon scientific workflow, so BenchLM does not use them as weighted base-model ranking inputs.
BenchLM mirrors the published rads average (pass@1) view for ResearchClawBench. Claude Code leads the public snapshot at 21.5% , followed by Claude Opus 4.8 (21.1%) and Claude Opus 4.7 (20.7%). BenchLM does not use these results to rank models overall.
Claude Code
InternScience
claude-opus-4-6
Claude Opus 4.8
Anthropic
claude-opus-4-8
Claude Opus 4.7
Anthropic
Claude-Opus-4.7
The published ResearchClawBench snapshot is tightly clustered at the top: Claude Code sits at 21.5%, while the third row is only 0.8 points behind. The broader top-10 spread is 3.3 points, so many of the published scores sit in a relatively narrow band.
28 models have been evaluated on ResearchClawBench. The benchmark falls in the Agentic category. This category carries a 22% weight in BenchLM.ai's overall scoring system. ResearchClawBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
Year
2026
Tasks
40 tasks across 10 scientific domains
Format
End-to-end autonomous research evaluation with RADS scoring
Difficulty
Scientific research re-discovery
ResearchClawBench grades scientific agents with RADS, a rubric where 50 indicates matching the target paper and 70+ indicates surpassing it. BenchLM mirrors the official Pass@1 leaderboard as display-only because rows reflect a research-agent harness and long-horizon scientific workflow, not a normalized base-model-only comparison.
Version
ResearchClawBench 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper.
Claude Code currently leads the published ResearchClawBench snapshot with 21.5% rads average (pass@1). BenchLM shows this benchmark for display only and does not use it in overall rankings.
28 AI models are included in BenchLM's mirrored ResearchClawBench snapshot, based on the public leaderboard captured on 2026-06-29 snapshot.
For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.