SEC-Bench Pro leaderboard
As of November 14, 2026, among the results listed here, Google Gemini 4 Carbon has the highest reported SEC-Bench Pro score, 88.2% (Pass@1), published by Google on October 28, 2026, ahead of Google Gemini 4 Argon at 83.4%. 10 models have a SEC-Bench Pro result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
10 models ranked · 4 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon | 88.2% | Pass@1 | October 28, 2026 · read from the maker's chart | |
| 2 | Gemini 4 Argon | 83.4% | Pass@1 | October 28, 2026 · read from the maker's chart | |
| 3 | GPT-5.6 Sol | 71.2% | Not published | OpenAI | August 7, 2025 (model release) |
| 4 | MiMo-V2.6-Pro | 66.3% | Not published | Xiaomi | September 21, 2026 (model release) |
| 5 | DeepSeek V4.1 Flash | 62.8% | Pass@1 | DeepSeek | September 10, 2026 (model release) |
| 6 | GPT-5.6 Terra | 57.7% | Not published | OpenAI | August 7, 2025 (model release) |
| 7 | GPT-5.6 Luna | 48.9% | Not published | OpenAI | August 7, 2025 (model release) |
| 8 | MiMo-V2.6-Flash | 47.5% | Not published | Xiaomi | September 21, 2026 (model release) |
| 9 | GPT-5.5 | 45.8% | Not published | OpenAI | August 7, 2025 (model release) |
| 10 | MiMo-V2.5-Pro | 17.7% | Not published | Xiaomi | September 21, 2026 (model release) |
Other reported runs
| Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|
| Gemini 4 Argon | 83.4% | Not published | October 28, 2026 · read from the maker's chart | |
| Gemini 4 Carbon | 88.2% | Not published | October 28, 2026 · read from the maker's chart |
How much to trust these numbers
- Source
- Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
- Settings
- Settings differ between makers. Check the Setting column before comparing two scores.
- Spread
- 88.2% to 17.7% across 10 models.
- Freshness
- Newest result published October 28, 2026; oldest August 7, 2025.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What SEC-Bench Pro measures
SEC-Bench Pro asks models to find real security vulnerabilities by reproducing working proof-of-concept inputs; its score reflects how well they perform this task. Source: SEC-bench Pro