MoE LLM Serving on MiniMax-M2 229B (FP8, 256 experts, top-8)
3.7Throughput (tok/s)WiSP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| WiSPSetting=CAP., GPU budget=one 94 GiB H100, cap amount=322026.06 | 3.7 | — |
Loading the SOTA2 catalog…
| Method | Links | ||
|---|---|---|---|
| WiSPSetting=CAP., GPU budget=one 94 GiB H100, cap amount=322026.06 | 3.7 | — |