Anthropic STEM Professor Bench (STEM Professor Bench)
We show this table for reference; we do not rank on it.
Anthropic's internal set of unpublished research problems written by professional scientists, each about a month of research work, mostly in theoretical physics; BenchLM stores the 64-problem subset.
Benchmark score on STEM Professor Bench — November 11, 2026
We compile the STEM Professor Bench rows from provider self-reports. Claude Fable 5.2 leads the table at 61.8%, followed by Claude Opus 5.5 (53.7%) and Claude Fable 5.1 (29.9%). We do not use these results to rank models overall.
Claude Fable 5.2
Anthropic
Claude Opus 5.5
Anthropic
Claude Fable 5.1
Anthropic
3 modelsReasoningCurrentDisplay onlyUpdated November 11, 2026
Benchmark score table (3 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Fable 5.2Anthropic | 61.8% | Not reported | Closed |
| 2 | Claude Opus 5.5Anthropic | 53.7% | Not reported | Closed |
| 3 | Claude Fable 5.1Anthropic | 29.9% | Not reported | Closed |
Among the reported STEM Professor Bench rows, Claude Fable 5.2 is first at 61.8%. The third row is 31.9 points behind. The broader top-10 range is 31.9 points, so the table still separates the published systems.
3 models have been evaluated on STEM Professor Bench. The benchmark falls in the Reasoning category. STEM Professor Bench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About STEM Professor Bench
Year
2026
Tasks
64 research problems
Format
Mean score over up to nine attempts
Difficulty
Research-level science
Each model works as a single agent with Python, bash and a file editor within each author's limits, with a fixed arXiv index up to December 1, 2025 on 58 problems. Claude Opus 5 grades 56 problems against answer keys or rubrics; the rest use author-written checkers. Values are read from the labeled bars of Figure 8.9.A. Internal, so display-only.
Freshness and provenance
Version
STEM Professor Bench 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.
Questions
What does STEM Professor Bench measure?
Anthropic's internal set of unpublished research problems written by professional scientists, each about a month of research work, mostly in theoretical physics; BenchLM stores the 64-problem subset.
Which model scores highest on STEM Professor Bench?
Claude Fable 5.2 by Anthropic currently leads with a score of 61.8% on STEM Professor Bench.
How many models are evaluated on STEM Professor Bench?
3 AI models have published results on STEM Professor Bench in the BenchLM catalog.
Compare top models on STEM Professor Bench
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.