ArmBench-LLM - NER: leaderboard
ArmBench-LLM category: Armenian named-entity recognition, the mean accuracy on the FiNER and PioNER tasks, evaluated with Metric AI Lab's lighteval fork. 0-100, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Not forecast. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Pro | 78.35 | #70 |
| 2 | Grok 4 Fast | 75.92 | #266 |
| 3 | Mistral Large 3 | 75.62 | #410 |
| 4 | Claude 3.7 Sonnet | 74.59 | #259 |
| 5 | Claude Sonnet 4 | 74.28 | #207 |
| 6 | Gemini 3 Flash (Preview) | 73.61 | #90 |
| 7 | Gemini 2.5 Flash | 73.6 | #255 |
| 8 | Grok 4.1 Fast | 73.52 | #227 |
| 9 | Qwen 3.5 27B | 72.46 | #192 |
| 10 | Gemma 4 31B (IT) | 71.69 | #178 |
| 11 | Gemma 3 12B (IT) | 70.2 | #672 |
| 12 | Grok 4.20 | 69.61 | #179 |
| 13 | Ministral 3 14B | 69.44 | #664 |
| 14 | Gemma 3 27B (IT) | 68.97 | #528 |
| 15 | GPT-5.4 Mini | 68.89 | #228 |
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-ner · How It Works · Data refreshed daily, snapshot 2026-11-05.