ArmBench-LLM v2 - Knowledge: leaderboard

ArmBench-LLM 5-category edition Knowledge score: mean across Armenian Unified Exams (History, Literature, Math rescaled from 0-20), INCLUDE, Hartak MCQA, Simple QA (BLEU/100) and Armenian MMLU-Pro. 0-100, higher is better.

Metric: Mean of exam (0-20 rescaled), accuracy and BLEU tasks (0-100). Source: huggingface.co. Saturation forecast: Around February 2027. 32 models tracked. Updated 22 Nov 2026.

Top models

#ModelScoreOverall rank
1GPT-6.1 Sol77.94#15
2DeepSeek V4.1 Flash70.94#69
3GPT-5.2 Pro69.52#139
4Gemini 3 Pro (Preview)64.71#123
5Gemini 3 Flash (Preview)63.91#156
6Grok 4 Fast63.05#446
7Claude 3.7 Sonnet59.72#455
8Grok 4.1 Fast59.08#431
9Qwen 3.5 27B58.22#380
10Gemini 3.1 Pro (Preview)56.09#102
11Claude Sonnet 454.21#414
12Gemma 4 31B (IT)54.17#354
13GPT-5.4 Mini53.47#409
14Gemini 2.5 Pro53.36#280
15Mistral Large 350.92#835

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=armbench-llm-v2-knowledge · How It Works · Data refreshed daily, snapshot 2026-11-22.