List your company

Language Modeling on FineWeb-Edu 20B tokens (val)

15.03Final PPLDense Softmax 1.3B (Exp 085)
14.8915.83516.7817.725May 10, 2026

Evaluation results

MethodLinks
Dense Softmax 1.3B (Exp 085)Total Parameters=1.31B, Active Parameters=1.31B, dff=8192, Wall time=53.3h2026.0515.03
MoL Hybrid 1+3of15 (Exp 087)Total Parameters=2.08B, Active Parameters=0.61B, dff=2048†, Wall time=102.0h2026.0518.04
Dense Softmax 0.7B (Exp 089)Total Parameters=0.71B, Active Parameters=0.71B, dff=2048, Wall time=44.5h2026.0518.53