List your company

Zero-shot Language Evaluation on 7-Task Suite (BoolQ, PIQA, HellaSwag, WinoGrande, ARC-E, ARC-C, OBQA) (test)

62.9BoolQ AccuracyMamba-2 FP16
55.72457.58759.4561.313Jun 16, 2026

Evaluation results

MethodLinks
Mamba-2 FP16Bits=162026.0662.973.759.961.160.433.237.455.5
Bi-Mamba 1.3BBits=1, Type=QAT scratch, Train tokens=105B2026.066068.847.355.94826.332.248.4
BitNet b1.58 1.3BBits=1.58, Type=QAT scratch, Train tokens=100B2026.0656.768.837.755.854.924.219.645.4
Ternary MambaBits=1.58, Type=QAT, Train tokens=102M2026.065669.347.556.348.225.933.848.1