TukaBench - English: leaderboard
The English condition of TukaBench (2026), a jailbreak benchmark for African languages: attack success rate on the English versions of its harmful prompts (JailbreakBench-derived, culturally adapted and African-authored), direct prompting, with a GPT-4.1 judge labelling each response jailbroken, refused or deflected. Scores are percentages; lower is better.
Metric: Attack success rate (%; share of harmful prompts answered Jailbroken) on the English versions of the TukaBench jailbreak prompts (JailbreakBench-derived Afri-JBB-Harm and Afri-JBB-Culture plus African-authored AfriJail-Mono), direct prompting, greedy decoding, GPT-4.1 judge labelling each response Jailbroken, Refused or Deflected; lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 2.1 | #94 |
| 2 | GPT-OSS-120B | 4.7 | #700 |
| 3 | Qwen 3.5 27B | 5.4 | #359 |
| 4 | GPT-5.2 | 5.8 | #176 |
| 5 | Claude Sonnet 5 | 9.4 | #111 |
| 6 | Claude Opus 4.8 | 10.9 | #56 |
| 7 | DeepSeek V3.2 | 11.4 | #406 |
| 8 | Claude Haiku 4.5 | 13.9 | #568 |
| 9 | Gemma 4 31B | 15.4 | #394 |
| 10 | GPT-4o | 16.5 | #741 |
| 11 | Llama 4 Maverick | 17.1 | #1099 |
| 12 | Grok 4.3 | 17.9 | #333 |
| 13 | Grok 4 | 23.1 | #307 |
| 14 | Gemma 3 27B | 26.1 | #1631 |
| 15 | Grok 3 | 32.8 | #602 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tukabench-english · How It Works · Data refreshed daily, snapshot 2026-11-13.