GB200 NVL72 FP4: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on GB200 NVL72 FP4 (NVIDIA Blackwell) running DeepSeek V4 Pro 1.6T. 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.
At 32 tok/s/user on DeepSeek V4 Pro 1.6T (GB200 NVL72 FP4), MTP delivers 9479 tok/s/GPU at $0.06 per million tokens; Off delivers 8088 tok/s/GPU at $0.08. MTP is 17% cheaper per token; MTP delivers 17% more tok/s/GPU. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution.
MTP posts 9479 tok/s/GPU for $0.06 per million tokens at 46 tok/s/user on DeepSeek V4 Pro 1.6T (GB200 NVL72 FP4); Off posts 2843 tok/s/GPU for $0.22. MTP is 237% cheaper per token; MTP delivers 233% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 61 tok/s/user on DeepSeek V4 Pro 1.6T (GB200 NVL72 FP4): MTP hits 9252 tok/s/GPU, Off hits 132. Per-million costs land at $0.07 and $11.6 respectively. MTP is 17397% cheaper per token; MTP delivers 6916% more tok/s/GPU. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size. (Numbers reflect this URL's pinned 8k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:9478.9Off:8087.6 | MTP:9478.9Off:2843.2 | MTP:9251.8Off:131.9 |
| Cost ($/M tok) | MTP:$0.065Off:$0.076 | MTP:$0.065Off:$0.218 | MTP:$0.066Off:$11.614 |
| tok/s/MW | MTP:4513744Off:3851256 | MTP:4513744Off:1353903 | MTP:4405600Off:62791 |
| Concurrency | MTP:~16384Off:~3693 | MTP:~16384Off:~364 | MTP:~12097Off:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.