B300 FP4: M3 EAGLE vs Off Speculative Decoding
Speculative decoding comparison of M3 EAGLE versus Off on B300 FP4 (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.
Near the low end of the 29–194 tok/s/user interactivity band, at 70 tok/s/user on MiniMax M3 428B (B300 FP4): M3 EAGLE runs 3494 tok/s/GPU at $0.19/M tokens, Off runs 1584 at $0.41/M. M3 EAGLE is 121% cheaper per token; M3 EAGLE delivers 121% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 112 tok/s/user on MiniMax M3 428B (B300 FP4), M3 EAGLE delivers 2237 tok/s/GPU at $0.29 per million tokens; Off delivers 807 tok/s/GPU at $0.81. M3 EAGLE is 179% cheaper per token; M3 EAGLE delivers 177% more tok/s/GPU. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution.
M3 EAGLE posts 1976 tok/s/GPU for $0.33 per million tokens at 153 tok/s/user on MiniMax M3 428B (B300 FP4); Off posts 246 tok/s/GPU for $2.57. M3 EAGLE is 682% cheaper per token; M3 EAGLE delivers 702% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level. (Numbers reflect this URL's pinned 1k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | M3 EAGLE:3494.3Off:1583.7 | M3 EAGLE:2237.0Off:807.2 | M3 EAGLE:1975.6Off:246.2 |
| Cost ($/M tok) | M3 EAGLE:$0.186Off:$0.412 | M3 EAGLE:$0.291Off:$0.812 | M3 EAGLE:$0.328Off:$2.567 |
| tok/s/MW | M3 EAGLE:1610277Off:729808 | M3 EAGLE:1030890Off:371999 | M3 EAGLE:910396Off:113474 |
| Concurrency | M3 EAGLE:~158Off:~123 | M3 EAGLE:~57Off:~7 | M3 EAGLE:~28Off:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.