ArmBench-LLM - MCQA: leaderboard
ArmBench-LLM category: Armenian multiple-choice question answering, the mean accuracy on the INCLUDE, Syndarin and Scientific question sets. 0-100, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Pro | 98.67 | #70 |
| 2 | Claude Sonnet 4 | 97.33 | #207 |
| 3 | Claude 3.7 Sonnet | 96.67 | #259 |
| 4 | Grok 4 Fast | 96 | #266 |
| 5 | GPT-5.4 Mini | 94.67 | #228 |
| 6 | Gemini 2.5 Pro | 94 | #159 |
| 7 | Qwen 3.5 27B | 91.33 | #192 |
| 8 | Gemini 3.1 Pro (Preview) | 90.67 | #64 |
| 9 | Gemini 3 Pro (Preview) | 90 | #71 |
| 10 | Gemini 3 Flash (Preview) | 88.67 | #90 |
| 11 | Gemma 4 31B (IT) | 87.33 | #178 |
| 12 | GLM-5 | 87.33 | #149 |
| 13 | GLM-4.7 Flash | 87.33 | #543 |
| 14 | Gemini 2.5 Flash | 86 | #255 |
| 15 | Grok 4.1 Fast | 86 | #227 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=armbench-llm-mcqa · How It Works · Data refreshed daily, snapshot 2026-11-05.