SEAL - Coding: leaderboard
Scale AI expert-evaluated coding benchmark with 1,000 prompts spanning debugging, optimization, documentation, and codebase analysis across multiple programming languages.
Metric: Score. Source: scale.com. Saturation forecast: Estimated already saturated. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 Mini | 1237 | #323 |
| 2 | O3 Mini | 1137 | #272 |
| 3 | GPT-4o | 1132 | #343 |
| 4 | O1 Preview | 1123 | #193 |
| 5 | Gemini 2.0 Flash | 1111 | #342 |
| 6 | Gemini 2.0 Flash (Thinking) | 1108 | #342 (Gemini 2.0 Flash) |
| 7 | DeepSeek R1 | 1100 | #251 |
| 8 | O1 (2024-12-17) | 1083 | #149 |
| 9 | Claude 3.5 Sonnet (20240620) | 1079 | #299 |
| 10 | GPT-4o (2024-08-06) | 1045 | #336 |
| 11 | GPT-4o (2024-05-13) | 1036 | #325 |
| 12 | GPT-4 Turbo (Preview) | 1034 | #397 |
| 13 | Mistral Large 2 (Jul) | 1029 | #484 |
| 14 | Llama 3.1 405B Instruct | 1022 | #457 |
| 15 | Gemini 1.5 Pro (0514) | 994 | #512 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seal-coding · How It Works · Data refreshed daily, snapshot 2026-10-20.