Enkokilish Bench (English): leaderboard
Enkokilish riddles translated to English (99 items): same contains-match scoring, isolating the Amharic language barrier from the riddle-solving ability itself.
Metric: Riddles Solved (%). Source: github.com. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Devstral Small 2 | 84.85 | #491 |
| 2 | DeepSeek V3.2 | 61.62 | #210 |
| 3 | Gemini 2.5 Flash | 60.61 | #255 |
| 4 | Gemini 2.5 Flash Lite | 60.61 | #432 |
| 5 | GPT-OSS-20B | 58.59 | #523 |
| 6 | Gemini 2.0 Flash Lite | 52.53 | #455 |
| 7 | OLMo-3.1 32B Think | 40.4 | #636 |
| 8 | Llama 3.1 8B | 34.34 | #1130 |
| 9 | Nova Lite | 30.3 | #741 |
| 10 | Grok 4 Fast (Non-reasoning) | 23.23 | #266 (Grok 4 Fast) |
| 11 | GLM 4.5 Air | 13.13 | #403 |
| 12 | Trinity-Mini | 9.09 | #767 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=enkokilish-bench-english · How It Works · Data refreshed daily, snapshot 2026-11-05.