B200 FP4: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on B200 FP4 (NVIDIA Blackwell) running DeepSeek R1. Throughput, cost, and interactivity differences across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
MTP acceptance-rate implementations differ across inference engines. Points from different engines are not directly comparable on the same curve — throughput and cost at matched interactivity may reflect engine-level differences rather than pure speculative decoding gains. Interpret cross-engine comparisons with caution.
MTP posts 8907 tok/s/GPU for $0.06 per million tokens at 62 tok/s/user on DeepSeek R1 (B200 FP4); Off posts 1518 tok/s/GPU for $0.37. MTP is 509% cheaper per token; MTP delivers 487% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 102 tok/s/user on DeepSeek R1 (B200 FP4): MTP hits 2795 tok/s/GPU, Off hits 483. Per-million costs land at $0.19 and $1.14 respectively. MTP is 488% cheaper per token; MTP delivers 478% more tok/s/GPU. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size.
Toward the upper edge of the 22–182 tok/s/user interactivity band, at 142 tok/s/user on DeepSeek R1 (B200 FP4): MTP runs 662 tok/s/GPU at $0.84/M tokens, Off runs 148 at $3.65/M. MTP is 335% cheaper per token; MTP delivers 346% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less. (Numbers reflect this URL's pinned 1k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:8906.5Off:1518.4 | MTP:2795.1Off:483.2 | MTP:661.6Off:148.4 |
| Cost ($/M tok) | MTP:$0.060Off:$0.366 | MTP:$0.193Off:$1.136 | MTP:$0.838Off:$3.650 |
| tok/s/MW | MTP:4104399Off:699737 | MTP:1288078Off:222662 | MTP:304889Off:68383 |
| Concurrency | MTP:~3578Off:~786 | MTP:~869Off:~53 | MTP:~113Off:~24 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.