Terminal-Bench 4.1 leaderboard
As of November 22, 2026, among the results listed here, Google Gemini 4 Hydrogen has the highest reported Terminal-Bench 4.1 score, 57.0% (Cheating-adjusted), published by Google on November 12, 2026, ahead of RadixArk DaisyRose 27B at 12.0%. 2 models have a Terminal-Bench 4.1 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
57.0%
Reported by Google, November 12, 2026 · Cheating-adjusted
2 models ranked · 2 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | Gemini 4 Hydrogen | 57.0% | Cheating-adjusted | November 12, 2026 · read from the maker's chart | |
| 2 | R DaisyRose 27B | 12.0% | Not published | Radixark | November 21, 2026 (model release) |
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
- 57.0% to 12.0% across 2 models.
- Freshness
- Newest result published November 21, 2026; oldest November 12, 2026.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What Terminal-Bench 4.1 measures
Terminal-Bench 4.1 tests whether AI agents can use a real terminal to complete software tasks; its score is the percentage of tasks resolved, with a 95% confidence interval. Source: TERMINAL-BENCH