List your company

Long-context capability evaluation on RULER 8192 length

93.75AccuracyQwen3-30B A3B-Instruct
78.742882.638986.53590.4311Feb 9, 2026

Evaluation results

MethodLinks
Qwen3-30B A3B-InstructBudget=Full2026.0293.75
QUOKAModel=Qwen3-30B A3B-Instruct, Budget=25%2026.0292.77
Qwen3-4BBudget=Full2026.0291.68
QUOKAModel=Qwen3-4B, Budget=25%2026.0291.35
Smollm3Budget=Full2026.0283.46
Qwen2.5-3BBudget=Full2026.0281.99
QUOKAModel=Smollm3, Budget=25%2026.0281.45
Llama3.2-3BBudget=Full2026.0281.33
QUOKAModel=GPT-OSS-20B, Budget=25%2026.0280.66
QUOKAModel=Qwen2.5-3B, Budget=25%2026.0279.78
QUOKAModel=Llama3.2-3B, Budget=25%2026.0279.72
GPT-OSS-20BBudget=Full2026.0279.32