TukaBench: leaderboard
Safety evaluation of Large Language Models (LLMs) remains heavily English-centric, leaving Low-Resource Languages (LRLs), particularly African ones, critically underexplored. We introduce TUKABENCH, a jailbreak benchmark for seven African languages that extends JailbreakBench (JBB) beyond direct translation through four settings: human translation of JBB prompts, English adaptation to African cont
Metric: ASR ↑ African Languages (self-reported). Source: benchmarklist.com. Saturation forecast: Around April 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 2.4 | #58 |
| 2 | Qwen 3.5 27B | 7.9 | #187 |
| 3 | Grok 4 | 9 | #181 |
| 4 | GPT-5.2 | 9.3 | #110 |
| 5 | GPT-OSS-120B | 10.7 | #341 |
| 6 | Grok 4.3 | 11.6 | #188 |
| 7 | Gemma 3 27B | 15.6 | #607 |
| 8 | Gemma 4 31B | 16.6 | #200 |
| 9 | DeepSeek V3.2 | 18.5 | #205 |
| 10 | Llama 4 Maverick | 21.9 | #464 |
| 11 | GPT-4o | 23.5 | #343 |
| 12 | Grok 3 | 31.5 | #304 |
| 13 | GPT-3.5 | 38.1 | #818 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=tukabench · How It Works · Data refreshed daily, snapshot 2026-10-20.