Speed test matrix
MiniMax-M2.1 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-m2.1
- 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-M2.1 --served-model-name minimax-m2.1 --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 context33.2K246 ms TTFT
16K context33.8K479 ms TTFT
32K context28.9K1,116 ms TTFT
64K context23K2,797 ms TTFT
128K context15.4K8,332 ms TTFT
187K context11.7K16,036 ms TTFT
192K context11.4K16,852 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 | 142.554 ms TTFT | 237.473 ms TTFT | 429.895 ms TTFT | 691.1144 ms TTFT | 1,127.4248 ms TTFT | Looped | 2,698.31,222 ms TTFT | Over capacity |
| 16K | 138.773 ms TTFT | 231.692 ms TTFT | 415.7133 ms TTFT | Looped | 1,046.5410 ms TTFT | 1,565.8797 ms TTFT | Over capacity | Over capacity |
| 32K | 133.1105 ms TTFT | Looped | 387.4215 ms TTFT | 593.1408 ms TTFT | Looped | Over capacity | Over capacity | Over capacity |
| 64K | 124.2154 ms TTFT | 201.2243 ms TTFT | 343.8426 ms TTFT | 512.3841 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 111.4282 ms TTFT | 172.8505 ms TTFT | 281.2908 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 187K | 101.2641 ms TTFT | 152.5834 ms TTFT | 242.01,589 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 | 33.2K | — | 246 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 33.8K | — | 479 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 28.9K | — | 1,116 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 23K | — | 2,797 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 15.4K | — | 8,332 ms | — | — | 1 | — | — |
| Prefill | 187K | 1 | Valid | 11.7K | — | 16,036 ms | — | — | 1 | — | — |
| Prefill | 192K | 1 | Valid | 11.4K | — | 16,852 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 142.5 | 143.1 | 54 ms | 7.0 ms | 1.0 | 1 | 4,274 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 237.4 | 123.5 | 73 ms | 8.1 ms | 2.0 | 2 | 7,120 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 429.8 | 107.7 | 95 ms | 9.3 ms | 4.0 | 0 | 12,892 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 691.1 | 87.0 | 144 ms | 11.5 ms | 8.0 | 0 | 20,728 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 1,127.4 | 71.1 | 248 ms | 14.1 ms | 16.0 | 0 | 33,824 | 30.0 s |
| Sustained decode | 8K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 64 | Valid | 2,698.3 | 42.8 | 1,222 ms | 23.4 ms | 64.0 | 0 | 80,959 | 30.0 s |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 138.7 | 139.4 | 73 ms | 7.2 ms | 1.0 | 1 | 4,161 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 231.6 | 115.9 | 92 ms | 8.6 ms | 2.0 | 0 | 6,951 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 415.7 | 104.1 | 133 ms | 9.6 ms | 4.0 | 0 | 12,470 | 30.0 s |
| Sustained decode | 16K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 16 | Valid | 1,046.5 | 66.0 | 410 ms | 15.2 ms | 16.0 | 0 | 31,392 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 1,565.8 | 49.6 | 797 ms | 20.2 ms | 32.0 | 0 | 46,976 | 30.0 s |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 133.1 | 133.9 | 105 ms | 7.5 ms | 1.0 | 1 | 3,992 | 30.0 s |
| Sustained decode | 32K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 4 | Valid | 387.4 | 97.1 | 215 ms | 10.3 ms | 4.0 | 0 | 11,621 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 593.1 | 74.7 | 408 ms | 13.4 ms | 8.0 | 0 | 17,792 | 30.0 s |
| Sustained decode | 32K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 124.2 | 126.2 | 154 ms | 7.9 ms | 1.0 | 1 | 3,725 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 201.2 | 100.8 | 243 ms | 9.9 ms | 2.0 | 0 | 6,036 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 343.8 | 86.4 | 426 ms | 11.6 ms | 4.0 | 0 | 10,320 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 512.3 | 64.9 | 841 ms | 15.4 ms | 8.0 | 0 | 15,368 | 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 | 282 ms | 9.0 ms | 1.0 | 0 | 3,341 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 172.8 | 86.8 | 505 ms | 11.5 ms | 2.0 | 0 | 5,184 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 281.2 | 71.0 | 908 ms | 14.1 ms | 4.0 | 0 | 8,436 | 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 | Valid | 101.2 | 101.2 | 641 ms | 9.9 ms | 1.0 | 0 | 3,033 | 30.0 s |
| Sustained decode | 187K | 2 | Valid | 152.5 | 76.8 | 834 ms | 13.0 ms | 2.0 | 0 | 4,576 | 30.0 s |
| Sustained decode | 187K | 4 | Valid | 242.0 | 61.6 | 1,589 ms | 16.2 ms | 4.0 | 0 | 7,259 | 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.