SWE-InfraBench: leaderboard
Metric: Correctness (%): share of tasks whose generated code passes all unit tests; 100 AWS CDK (Python) infrastructure-as-code tasks built from 34 open-source and custom repositories: the model writes a masked code block from a natural-language request inside a real project, and the synthesized CloudFormation template is checked by hidden unit tests; one attempt per task, temperature 0.25, Anthropic models prompted with extra XML tags; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.7 Sonnet | 34 |
| 2 | Claude 3.5 Sonnet (20241022) | 32 |
| 3 | Claude 3.5 Sonnet (20240620) | 29 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 29 |
| 5 | Gemini 2.5 Pro (Preview 05-06) | 29 |
| 6 | DeepSeek R1 | 24 |
| 7 | O3 | 23 |
| 8 | O4 Mini | 23 |
| 9 | GPT-4.1 | 18 |
| 10 | Llama 3.1 405B Instruct | 9 |
| 11 | Claude 3 Haiku | 8 |
| 12 | Llama 4 Maverick Instruct | 8 |
| 13 | Gemini 2.0 Flash Lite | 5 |
| 14 | GPT-4o Mini | 4 |
| 15 | Llama 4 Scout Instruct | 2 |
Interactive version: theaggregate.ai/benchmark?slug=swe-infrabench · How It Works · Data refreshed daily, snapshot 2026-09-29.