B200 FP8: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on B200 FP8 (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 5160 tok/s/GPU for $0.11 per million tokens at 44 tok/s/user on DeepSeek R1 (B200 FP8); Off posts 1713 tok/s/GPU for $0.32. MTP is 198% cheaper per token; MTP delivers 201% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 75 tok/s/user on DeepSeek R1 (B200 FP8): MTP hits 2558 tok/s/GPU, Off hits 1052. Per-million costs land at $0.23 and $0.51 respectively. MTP is 127% cheaper per token; MTP delivers 143% 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 13–137 tok/s/user interactivity band, at 106 tok/s/user on DeepSeek R1 (B200 FP8): MTP runs 1313 tok/s/GPU at $0.41/M tokens, Off runs 503 at $1.09/M. MTP is 164% cheaper per token; MTP delivers 161% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less. (Numbers reflect this URL's pinned 8k/1k · fp8 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:5160.2Off:1713.3 | MTP:2557.7Off:1052.2 | MTP:1313.1Off:502.9 |
| Cost ($/M tok) | MTP:$0.106Off:$0.316 | MTP:$0.226Off:$0.512 | MTP:$0.413Off:$1.090 |
| tok/s/MW | MTP:2377969Off:789521 | MTP:1178668Off:484863 | MTP:605123Off:231747 |
| Concurrency | MTP:~467Off:~98 | MTP:~160Off:~62 | MTP:~87Off:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.