Speed test matrix
MiniMax-M2 on H200 SXM 141GB ×4
sglang · 0.5.10.post1 · FP86 contexts (8K–187K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 187K, 192K
Hardware
- Type
- Discrete GPU
- GPU
- H200 SXM
- GPU count
- 4
- VRAM
- 564 GB (4 × 141 GB)
- OS
- linux
Engine
- Engine
- sglang
- Version
- 0.5.10.post1
- Quantization
- FP8
- Backend
- cuda
- Repository
- github.com/sgl-project/sglang
- Build
- cuda-13.2
Run
- Served model
- minimax-m2
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.6.2
- Decode test
- Fixed duration
- Prefill mode
- Measured during decode warm-up
- Time per measurement
- 30 s
- Max output
- 4,096 tokens
Launch command
"$HOME/env_h200-minimax-m2-sglang-v0510-post1/bin/sglang" serve --model-path MiniMaxAI/MiniMax-M2 --served-model-name minimax-m2 --trust-remote-code --tp 4 --reasoning-parser minimax --tool-call-parser minimax-m2 --context-length 196608 --mem-fraction-static 0.839155 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --dtype bfloat16
Prefill
prompt tok/s by context length8K context33.4K245 ms TTFT
16K context33.5K484 ms TTFT
32K context29.3K1,101 ms TTFT
64K context23K2,800 ms TTFT
128K context15.4K8,337 ms TTFT
187K context11.7K15,979 ms TTFT
192K context11.4K16,871 ms TTFT
Sustained decode
total tok/s · rows are context length, columns are simultaneous requests| Context | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128 |
|---|---|---|---|---|---|---|---|---|
| 8K | 142.650 ms TTFT | 237.7231 ms TTFT | 429.0100 ms TTFT | 737.9143 ms TTFT | 1,184.8248 ms TTFT | 1,822.8501 ms TTFT | 2,744.71,178 ms TTFT | Over capacity |
| 16K | 138.770 ms TTFT | 230.792 ms TTFT | Looped | 736.8220 ms TTFT | Looped | Looped | Over capacity | Over capacity |
| 32K | 133.195 ms TTFT | 219.9135 ms TTFT | 403.1225 ms TTFT | 597.0409 ms TTFT | 921.8789 ms TTFT | Over capacity | Over capacity | Over capacity |
| 64K | 124.0156 ms TTFT | 200.9245 ms TTFT | 346.7426 ms TTFT | 508.2843 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 111.4286 ms TTFT | Looped | 281.4911 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 187K | Looped | 152.9854 ms TTFT | 241.71,612 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. Over capacity: not run because this many requests at this context would exceed the server's KV cache.
TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.
All measurements
Latencies are medians. Total tok/s is output across all simultaneous requests.| Phase | Context | Requests at once | Status | Total tok/s | Per request tok/s | TTFT | Time between tokens | Average active requests | Completed requests | Output tokens | Measured for |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Prefill | 8K | 1 | Valid | 33.4K | — | 245 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 33.5K | — | 484 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 29.3K | — | 1,101 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 23K | — | 2,800 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 15.4K | — | 8,337 ms | — | — | 1 | — | — |
| Prefill | 187K | 1 | Valid | 11.7K | — | 15,979 ms | — | — | 1 | — | — |
| Prefill | 192K | 1 | Valid | 11.4K | — | 16,871 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 142.6 | 143.1 | 50 ms | 7.0 ms | 1.0 | 1 | 4,277 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 237.7 | 118.0 | 231 ms | 8.5 ms | 2.0 | 0 | 7,130 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 429.0 | 107.6 | 100 ms | 9.3 ms | 4.0 | 0 | 12,869 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 737.9 | 92.0 | 143 ms | 10.9 ms | 8.0 | 0 | 22,135 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 1,184.8 | 73.7 | 248 ms | 13.6 ms | 16.0 | 0 | 35,546 | 30.0 s |
| Sustained decode | 8K | 32 | Valid | 1,822.8 | 57.5 | 501 ms | 17.4 ms | 32.0 | 0 | 54,679 | 30.0 s |
| Sustained decode | 8K | 64 | Valid | 2,744.7 | 43.9 | 1,178 ms | 22.8 ms | 64.0 | 0 | 82,356 | 30.0 s |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 138.7 | 139.4 | 70 ms | 7.2 ms | 1.0 | 1 | 4,160 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 230.7 | 115.6 | 92 ms | 8.7 ms | 2.0 | 0 | 6,921 | 30.0 s |
| Sustained decode | 16K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 8 | Valid | 736.8 | 91.4 | 220 ms | 10.9 ms | 8.0 | 0 | 22,101 | 30.0 s |
| Sustained decode | 16K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 133.1 | 133.9 | 95 ms | 7.5 ms | 1.0 | 1 | 3,993 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 219.9 | 110.2 | 135 ms | 9.1 ms | 2.0 | 0 | 6,598 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 403.1 | 100.5 | 225 ms | 9.9 ms | 4.0 | 0 | 12,094 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 597.0 | 75.2 | 409 ms | 13.3 ms | 8.0 | 0 | 17,904 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 921.8 | 58.3 | 789 ms | 17.2 ms | 16.0 | 0 | 27,645 | 30.0 s |
| Sustained decode | 32K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 124.0 | 126.1 | 156 ms | 7.9 ms | 1.0 | 1 | 3,721 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 200.9 | 100.7 | 245 ms | 9.9 ms | 2.0 | 0 | 6,025 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 346.7 | 87.0 | 426 ms | 11.5 ms | 4.0 | 0 | 10,398 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 508.2 | 64.4 | 843 ms | 15.5 ms | 8.0 | 0 | 15,247 | 30.0 s |
| Sustained decode | 64K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 111.4 | 111.5 | 286 ms | 9.0 ms | 1.0 | 0 | 3,343 | 30.0 s |
| Sustained decode | 128K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 4 | Valid | 281.4 | 71.0 | 911 ms | 14.1 ms | 4.0 | 0 | 8,442 | 30.0 s |
| Sustained decode | 128K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 187K | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 187K | 2 | Valid | 152.9 | 77.0 | 854 ms | 13.0 ms | 2.0 | 0 | 4,588 | 30.0 s |
| Sustained decode | 187K | 4 | Valid | 241.7 | 61.5 | 1,612 ms | 16.3 ms | 4.0 | 0 | 7,256 | 30.0 s |
| Sustained decode | 187K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 187K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 187K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 187K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 187K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. Over capacity: not run because this many requests at this context would exceed the server's KV cache.