ArmBench-LLM v2 - Generation: leaderboard

ArmBench-LLM 5-category edition Generation score: mean BLEU across Armenian email summarization, conversation summarization, paraphrase and English-Armenian short-text translation. 0-100 BLEU, higher is better.

Metric: BLEU (0-100). Source: huggingface.co. Saturation forecast: Around 2034. 32 models tracked. Updated 22 Nov 2026.

Top models

#ModelScoreOverall rank
1Ayb-0.5-9B27
2Ayb-0.5-4B24.85
3Gemini 2.5 Flash18.95#466
4Ayb-0.5-0.8B18.92
5Claude Sonnet 418.7#414
6Gemma 4 31B (IT)17.71#354
7Claude 3.7 Sonnet17.32#455
8GPT-5.4 Mini16.99#409
9Gemini 3 Flash (Preview)16.32#156
10Grok 4.1 Fast15.64#431
11GPT-5.2 Pro15.6#139
12GPT-5.4 Nano15.25#653
13GPT-6.1 Sol15.2#15
14DeepSeek V4.1 Flash15.08#69
15Grok 4 Fast14.73#446

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-generation · How It Works · Data refreshed daily, snapshot 2026-11-22.