Kaggle Open Benchmarks - SciCode Subproblems: leaderboard
SciCode subproblem pass rate: the share of scientific-computing subproblems (drawn from 65 research coding problems across physics, chemistry, biology, mathematics and materials science) whose generated code passes the tests, in the standard setting without added background text. Kaggle's standardized re-run on its own harness under the open-benchmarks organization, not the original authors' leaderboard; Kaggle's fraction is shown as a percentage (0-100, higher is better).
Metric: Subproblem Pass Rate (%). Source: www.kaggle.com. Saturation forecast: Around 2029. 60 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 52.23 | #3 |
| 2 | Gemini 4 Argon | 45.02 | #9 |
| 3 | Gemini 3.1 Pro (Preview) | 44.9 | #64 |
| 4 | GPT-6 Astra | 43.3 | #6 |
| 5 | Claude Sonnet 5.5 | 43.3 | #14 |
| 6 | Gemini 3.7 Flash | 42.96 | #24 |
| 7 | GPT-5.6 Sol | 42.61 | #22 |
| 8 | Claude Opus 5 | 42.61 | #16 |
| 9 | Grok 4.5 | 41.9 | #54 |
| 10 | Gemini 3.8 Flash | 40.21 | #30 |
| 11 | Claude Sonnet 5 | 40.1 | #65 |
| 12 | Gemini 3.5 Flash | 39.52 | #66 |
| 13 | GPT-5.6 Terra | 39.18 | #40 |
| 14 | Claude Opus 4.8 | 38.14 | #39 |
| 15 | Gemini 3.6 Flash | 37.8 | #57 |
No result here: #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-open-benchmarks-scicode-subproblems · How It Works · Data refreshed daily, snapshot 2026-11-05.