CWE-bench leaderboard
As of November 14, 2026, among the results listed here, Google Gemini 4 Carbon has the highest reported CWE-bench score, 71.0% (Pass@1 (ties broken by Pass@4); Antigravity), published by Google on October 28, 2026, ahead of Google Gemini 4 Argon at 68.0%. 2 models have a CWE-bench result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
Best reported score
71.0%
Reported by Google, October 28, 2026 · Pass@1 (ties broken by Pass@4); Antigravity
Runner-up
68.0%
Reported by Google, October 28, 2026 · Pass@1 (ties broken by Pass@4); Antigravity
2 models ranked · 1 maker
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon | 71.0% | Pass@1 (ties broken by Pass@4); Antigravity | October 28, 2026 · read from the maker's chart | |
| 2 | Gemini 4 Argon | 68.0% | Pass@1 (ties broken by Pass@4); Antigravity | October 28, 2026 · read from the maker's chart |
Other reported runs
| Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|
| Gemini 4 Argon | 68.0% | Not published | October 28, 2026 · read from the maker's chart | |
| Gemini 4 Carbon | 71.0% | 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
- All ranked results use the same reported setting.
- Spread
- 71.0% to 68.0% across 2 models.
- Freshness
- Published October 28, 2026.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What CWE-bench measures
Tests models on auditing code repositories and repairing security weaknesses; the provided results do not explain what the score means. Source: CWE-bench v2: a cybersecurity benchmark by Collinear AI