Language Modeling on FineWeb-Edu (Precision-Specific PPL and Training Throughput)
8.49Perplexity (BF16)
Hyperloop
Updated 1mo ago
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
2026.04 | 8.49 | 8.59 | 180,000 | |
2026.04 | 8.57 | 8.62 | 109,000 | |
2026.04 | 8.6 | 8.71 | 181,000 | |
2026.04 | 8.68 | 8.97 | 183,000 | |
2026.04 | 9.65 | 9.81 | 354,000 | |
2026.04 | 10.02 | 10.24 | 367,000 | |
2026.04 | 10.07 | 10.16 | 237,000 | |
2026.04 | 10.19 | 10.27 | 367,000 | |
2026.04 | 14.4 | 14.68 | 750,000 | |
2026.04 | 14.55 | 14.73 | 514,000 | |
2026.04 | 14.65 | 14.85 | 786,000 | |
2026.04 | 14.85 | 15.18 | 786,000 |