List your company

Language Model Training on FineWeb (val)

95.2Training Time (s)Aster's Solution
95.13695.5689696.432Feb 3, 2026

Evaluation results

MethodLinks
Aster's SolutionHardware=8 NVIDIA H100 GPUs, Target Metric=less than 3.28 cross-entropy loss2026.0295.21.6
Previous BestHardware=8 NVIDIA H100 GPUs, Target Metric=less than 3.28 cross-entropy loss2026.0296.8—