B200 FP8: M3 EAGLE vs Off Speculative Decoding
Speculative decoding comparison of M3 EAGLE versus Off on B200 FP8 (NVIDIA Blackwell) running MiniMax M3 428B. 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.
Throughput at 59 tok/s/user on MiniMax M3 428B (B200 FP8): M3 EAGLE hits 2593 tok/s/GPU, Off hits 1317. Per-million costs land at $0.21 and $0.41 respectively. M3 EAGLE is 99% cheaper per token; M3 EAGLE delivers 97% 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.
Around the middle of the 21–174 tok/s/user interactivity band, at 97 tok/s/user on MiniMax M3 428B (B200 FP8): M3 EAGLE runs 1500 tok/s/GPU at $0.38/M tokens, Off runs 640 at $0.84/M. M3 EAGLE is 123% cheaper per token; M3 EAGLE delivers 134% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 136 tok/s/user on MiniMax M3 428B (B200 FP8), M3 EAGLE delivers 1271 tok/s/GPU at $0.45 per million tokens; Off delivers 226 tok/s/GPU at $2.41. M3 EAGLE is 440% cheaper per token; M3 EAGLE delivers 464% more tok/s/GPU. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution. (Numbers reflect this URL's pinned 1k/1k · fp8 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | M3 EAGLE:2592.9Off:1316.8 | M3 EAGLE:1500.4Off:640.1 | M3 EAGLE:1271.1Off:225.5 |
| Cost ($/M tok) | M3 EAGLE:$0.208Off:$0.414 | M3 EAGLE:$0.378Off:$0.844 | M3 EAGLE:$0.446Off:$2.406 |
| tok/s/MW | M3 EAGLE:1194880Off:606823 | M3 EAGLE:691442Off:294995 | M3 EAGLE:585762Off:103917 |
| Concurrency | M3 EAGLE:~128Off:~44 | M3 EAGLE:~44Off:~13 | M3 EAGLE:~18Off:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.