Speed test matrix
MiniMax-M2.7 on H200 SXM 141GB ×4
sglang · 0.5.10.post1 · FP86 contexts (8K–195K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 195K, 200K
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-m2.7
- 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-M2.7 --served-model-name minimax-m2.7 --trust-remote-code --tp 4 --reasoning-parser minimax --tool-call-parser minimax-m2 --context-length 204800 --mem-fraction-static 0.839212 --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 context34K476 ms TTFT
32K context29.9K1,081 ms TTFT
64K context23K2,797 ms TTFT
128K context15.4K8,332 ms TTFT
195K context11.3K17,287 ms TTFT
200K context11.2K17,972 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 | 142.152 ms TTFT | 237.366 ms TTFT | 428.875 ms TTFT | 711.6140 ms TTFT | 1,161.8248 ms TTFT | 1,798.4661 ms TTFT | 2,726.2887 ms TTFT | Over capacity |
| 16K | 138.393 ms TTFT | 230.992 ms TTFT | 420.2132 ms TTFT | 661.5233 ms TTFT | 1,079.6415 ms TTFT | 1,642.1838 ms TTFT | Over capacity | Over capacity |
| 32K | 133.296 ms TTFT | 226.2140 ms TTFT | 391.7223 ms TTFT | 629.1427 ms TTFT | 937.9823 ms TTFT | Over capacity | Over capacity | Over capacity |
| 64K | 124.4153 ms TTFT | 201.4250 ms TTFT | 352.0424 ms TTFT | Looped | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 111.5282 ms TTFT | 172.7494 ms TTFT | 282.8980 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 195K | 99.8501 ms TTFT | 153.21,020 ms TTFT | 236.71,719 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 | 34K | — | 476 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 29.9K | — | 1,081 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 23K | — | 2,797 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 15.4K | — | 8,332 ms | — | — | 1 | — | — |
| Prefill | 195K | 1 | Valid | 11.3K | — | 17,287 ms | — | — | 1 | — | — |
| Prefill | 200K | 1 | Valid | 11.2K | — | 17,972 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 142.1 | 142.9 | 52 ms | 7.0 ms | 1.0 | 1 | 4,263 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 237.3 | 120.7 | 66 ms | 8.3 ms | 2.0 | 2 | 7,117 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 428.8 | 107.5 | 75 ms | 9.3 ms | 4.0 | 0 | 12,863 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 711.6 | 89.0 | 140 ms | 11.2 ms | 8.0 | 0 | 21,343 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 1,161.8 | 72.9 | 248 ms | 13.7 ms | 16.0 | 0 | 34,845 | 30.0 s |
| Sustained decode | 8K | 32 | Valid | 1,798.4 | 56.7 | 661 ms | 17.6 ms | 32.0 | 0 | 53,945 | 30.0 s |
| Sustained decode | 8K | 64 | Valid | 2,726.2 | 43.3 | 887 ms | 23.1 ms | 64.0 | 0 | 81,778 | 30.0 s |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 138.3 | 139.6 | 93 ms | 7.2 ms | 1.0 | 1 | 4,147 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 230.9 | 115.6 | 92 ms | 8.6 ms | 2.0 | 0 | 6,928 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 420.2 | 105.1 | 132 ms | 9.5 ms | 4.0 | 0 | 12,604 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 661.5 | 83.1 | 233 ms | 12.0 ms | 8.0 | 0 | 19,839 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 1,079.6 | 67.8 | 415 ms | 14.7 ms | 16.0 | 0 | 32,381 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 1,642.1 | 51.6 | 838 ms | 19.4 ms | 32.0 | 0 | 49,242 | 30.0 s |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 133.2 | 134.0 | 96 ms | 7.5 ms | 1.0 | 1 | 3,995 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 226.2 | 112.9 | 140 ms | 8.9 ms | 2.0 | 0 | 6,786 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 391.7 | 98.0 | 223 ms | 10.2 ms | 4.0 | 0 | 11,751 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 629.1 | 78.7 | 427 ms | 12.7 ms | 8.0 | 0 | 18,871 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 937.9 | 59.2 | 823 ms | 16.9 ms | 16.0 | 0 | 28,138 | 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 | 124.4 | 126.1 | 153 ms | 7.9 ms | 1.0 | 1 | 3,732 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 201.4 | 100.9 | 250 ms | 9.9 ms | 2.0 | 0 | 6,042 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 352.0 | 88.1 | 424 ms | 11.3 ms | 4.0 | 0 | 10,561 | 30.0 s |
| Sustained decode | 64K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | 111.5 | 111.6 | 282 ms | 9.0 ms | 1.0 | 0 | 3,346 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 172.7 | 86.7 | 494 ms | 11.5 ms | 2.0 | 0 | 5,181 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 282.8 | 71.3 | 980 ms | 14.0 ms | 4.0 | 0 | 8,484 | 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 | 195K | 1 | Valid | 99.8 | 99.9 | 501 ms | 10.0 ms | 1.0 | 0 | 2,995 | 30.0 s |
| Sustained decode | 195K | 2 | Valid | 153.2 | 77.0 | 1,020 ms | 13.0 ms | 2.0 | 0 | 4,594 | 30.0 s |
| Sustained decode | 195K | 4 | Valid | 236.7 | 60.3 | 1,719 ms | 16.6 ms | 4.0 | 0 | 7,099 | 30.0 s |
| Sustained decode | 195K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 195K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 195K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 195K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 195K | 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.