GSO-Bench
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues.
The Frontier
Best score over time · one chart, every benchmark
Full rankings
18 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 | 33.3 | |
| 2 | 27.4 | |
| 3 | 26.5 | |
| 4 | 18.6 | |
| 5 | 14.7 | |
| 6 | 13.7 | |
| 7 | 9.8 | |
| 8 | 8.8 | |
| 9 | 6.9 | |
| 10 | 6.9 | |
| 11 | 4.9 | |
| 12 | 4.9 | |
| 13 | 4.6 | |
| 14 | 3.9 | |
| 15 | 3.8 | |
| 16 | 3.6 | |
| 17 | 1.3 | |
| 18 | 0.1 |
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with GSO-Bench
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About GSO-Bench
What does GSO-Bench measure?
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues. 18 AI models have been tested on it. Scores range from 0.1 to 33.3 out of 100.
Which model leads on GSO-Bench?
Claude Opus 4.6 from Anthropic leads GSO-Bench with a score of 33.3. The median score across 18 tested models is 6.9.
Is GSO-Bench saturated?
No · the top score is 33.3 out of 100 (33%). There is still meaningful room for improvement on GSO-Bench.
Does GSO-Bench predict performance on other benchmarks?
Yes · GSO-Bench scores correlate 0.98 with Cybench across 9 shared models. Models that do well on GSO-Bench tend to do well on Cybench.
How often is GSO-Bench data refreshed?
BenchGecko pulls updates daily. New model scores on GSO-Bench appear as soon as they are published by Epoch AI or the model provider.
- Category
- Code
- Max score
- 100
- Models
- 18
- Updated
- 2026-02-04
Top on GSO-Bench
Claude Opus 4.6 · 33.3GPT-5.2 · 27.4Claude Opus 4.5 · 26.5Gemini 3 Pro · 18.6Claude Sonnet 4.5 · 14.7More code benchmarks
Same category · related evaluations