| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| GPT2-1.5B | Throughput4.1 | 25 | 3mo ago | ||
| Open-PerfectBlend | SpecForge | Step Time (s)0.52 | 9 | 2mo ago | |
| Qwen3-30B-A3B workload | PithTrain | Throughput (tokens/s)280,000 | 7 | 2d ago | |
| Qwen3 32B (train) | Ulysses | Training Throughput (128K Seq Len)545.29 | 5 | 3mo ago | |
| Llama3 8B (train) | Ulysses | Throughput (128K SeqLen)2,320.47 | 5 | 3mo ago | |
| DeepSeek-V2-Lite workload | PithTrain | Training Throughput (tokens/s)114,600 | 3 | 2d ago | |
| GPT-OSS-20B workload | PithTrain | Throughput (tokens/s)140,900 | 2 | 2d ago |
