ArmBench-LLM v2 - Analysis: leaderboard
ArmBench-LLM 5-category edition Analysis score: mean accuracy across Armenian Topic-14 classification, sentiment analysis, PioNER, FiNER named-entity recognition and part-of-speech tagging. 0-100, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around October 2027. 32 models tracked. Updated 22 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6.1 Sol | 71.24 | #15 |
| 2 | DeepSeek V4.1 Flash | 70.98 | #69 |
| 3 | Gemma 4 31B (IT) | 70.39 | #354 |
| 4 | Claude 3.7 Sonnet | 70.01 | #455 |
| 5 | Gemini 3 Flash (Preview) | 69.96 | #156 |
| 6 | Claude Sonnet 4 | 69.04 | #414 |
| 7 | Gemini 2.5 Flash | 68.45 | #466 |
| 8 | Qwen 3.5 27B | 68.27 | #380 |
| 9 | Grok 4 Fast | 66.67 | #446 |
| 10 | Ayb-0.5-9B | 65.26 | |
| 11 | Ayb-0.5-4B | 65.2 | |
| 12 | Grok 4.1 Fast | 63.31 | #431 |
| 13 | Gemini 3 Pro (Preview) | 61.65 | #123 |
| 14 | Gemini 2.5 Pro | 60.58 | #280 |
| 15 | GPT-5.4 Mini | 60.46 | #409 |
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-analysis · How It Works · Data refreshed daily, snapshot 2026-11-22.