Share your thoughts, 1 month free Claude Pro on usSee more

Long-context Retrieval on RULER 32k context (test)

91.7Accuracy

Ministral-3-8B-Instruct-2512-BF16

Updated 2mo ago

Evaluation Results

Method	Links
Ministral-3-8B-Instruct-2512-BF16 2026.05		91.7
Qwen3-8B 2026.05		91
Qwen3-4B 2026.05		89
GPT-5 nano 2026.05		88.9
Llama-3.1-8B-Instruct 2026.05		87.3
gemma-3-12b-it 2026.05		80.4
Llama-3.2-3B-Instruct 2026.05		77.8
gpt-oss-20b 2026.05		77.1
gemma-3-4b-it 2026.05		62.3
EngGPT2-16B-A3B 2026.05		42.6
FastwebMIIA-7B 2026.05		33.9
Moonlight-16B-A3B-Instruct 2026.05		32.7
LLaMAntino-3-ANITA-8B-Inst-DPO-ITA 2026.05		28.5
deepseek-moe-16b-chat 2026.05		16.9
Minerva-7B-instruct-v1,0 2026.05		10.1
Velvet-14B 2026.05		0