Speed test matrix
Olmo-3.1-32B-Instruct-DPO on H200 NVL 141GB
vllm · 0.26.1rc1.dev638+g52be12cfa · BF164 contexts (8K–59K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 59K, 64K
Hardware
- Type
- Discrete GPU
- GPU
- H200 NVL
- VRAM
- 141 GB
- OS
- linux
Engine
- Engine
- vllm
- Version
- 0.26.1rc1.dev638+g52be12cfa
- Quantization
- BF16
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- cuda-13.2
- Commit
52be12cfac0c
Run
- Served model
- olmo
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.4.29
- 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-allenai-vllm/bin/vllm" serve allenai/Olmo-3.1-32B-Instruct-DPO --served-model-name olmo --trust-remote-code --tensor-parallel-size 1 --max-model-len 65536 --gpu-memory-utilization 0.871691 --host 0.0.0.0 --port 8000 --api-key <redacted>
Prefill
prompt tok/s by context length8K context8,762614 ms TTFT
16K context8,0771,311 ms TTFT
32K context7,9332,649 ms TTFT
59K context7,6205,066 ms TTFT
64K context7,5655,534 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 | 50.349 ms TTFT | 99.275 ms TTFT | 193.3104 ms TTFT | 365.6144 ms TTFT | 665.9217 ms TTFT | 1,119.5322 ms TTFT | 1,640.0539 ms TTFT | Over capacity |
| 16K | 50.178 ms TTFT | 98.889 ms TTFT | 191.3135 ms TTFT | 358.5195 ms TTFT | 639.6304 ms TTFT | 1,043.5356 ms TTFT | Over capacity | Over capacity |
| 32K | 49.786 ms TTFT | 97.7129 ms TTFT | 187.3156 ms TTFT | 345.3282 ms TTFT | 601.5324 ms TTFT | Over capacity | Over capacity | Over capacity |
| 59K | 49.0113 ms TTFT | 95.7150 ms TTFT | 181.0306 ms TTFT | 325.4386 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
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 | 8,762.0 | — | 614 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 8,077.0 | — | 1,311 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 7,933.0 | — | 2,649 ms | — | — | 1 | — | — |
| Prefill | 59K | 1 | Valid | 7,620.0 | — | 5,066 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 7,565.0 | — | 5,534 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 50.3 | 50.3 | 49 ms | 19.9 ms | 1.0 | 0 | 1,510 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 99.2 | 49.7 | 75 ms | 20.1 ms | 2.0 | 0 | 2,976 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 193.3 | 48.3 | 104 ms | 20.7 ms | 4.0 | 0 | 5,796 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 365.6 | 45.7 | 144 ms | 21.9 ms | 8.0 | 0 | 10,960 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 665.9 | 41.7 | 217 ms | 24.0 ms | 16.0 | 0 | 19,968 | 30.0 s |
| Sustained decode | 8K | 32 | Valid | 1,119.5 | 35.1 | 322 ms | 28.5 ms | 32.0 | 0 | 33,568 | 30.0 s |
| Sustained decode | 8K | 64 | Valid | 1,640.0 | 25.8 | 539 ms | 38.8 ms | 64.0 | 0 | 49,152 | 30.0 s |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 50.1 | 50.1 | 78 ms | 20.0 ms | 1.0 | 0 | 1,503 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 98.8 | 49.4 | 89 ms | 20.2 ms | 2.0 | 0 | 2,964 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 191.3 | 47.8 | 135 ms | 20.9 ms | 4.0 | 0 | 5,736 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 358.5 | 44.8 | 195 ms | 22.3 ms | 8.0 | 0 | 10,752 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 639.6 | 40.1 | 304 ms | 25.0 ms | 16.0 | 0 | 19,184 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 1,043.5 | 32.7 | 356 ms | 30.6 ms | 32.0 | 0 | 31,296 | 30.0 s |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 49.7 | 49.7 | 86 ms | 20.1 ms | 1.0 | 0 | 1,490 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 97.7 | 48.9 | 129 ms | 20.5 ms | 2.0 | 0 | 2,930 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 187.3 | 46.8 | 156 ms | 21.3 ms | 4.0 | 0 | 5,616 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 345.3 | 43.2 | 282 ms | 23.1 ms | 8.0 | 0 | 10,360 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 601.5 | 37.6 | 324 ms | 26.6 ms | 16.0 | 0 | 18,032 | 30.0 s |
| Sustained decode | 32K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 59K | 1 | Valid | 49.0 | 49.0 | 113 ms | 20.4 ms | 1.0 | 0 | 1,470 | 30.0 s |
| Sustained decode | 59K | 2 | Valid | 95.7 | 47.8 | 150 ms | 20.9 ms | 2.0 | 0 | 2,868 | 30.0 s |
| Sustained decode | 59K | 4 | Valid | 181.0 | 45.3 | 306 ms | 22.1 ms | 4.0 | 0 | 5,428 | 30.0 s |
| Sustained decode | 59K | 8 | Valid | 325.4 | 40.7 | 386 ms | 24.6 ms | 8.0 | 0 | 9,760 | 30.0 s |
| Sustained decode | 59K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 59K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 59K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 59K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.