H100 FP8: M3 EAGLE vs Off Speculative Decoding
Speculative decoding comparison of M3 EAGLE versus Off on H100 FP8 (NVIDIA Hopper) 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 33–144 tok/s/user interactivity band, at 60 tok/s/user on MiniMax M3 428B (H100 FP8): M3 EAGLE runs 737 tok/s/GPU at $0.49/M tokens, Off runs 278 at $1.29/M. M3 EAGLE is 161% cheaper per token; M3 EAGLE delivers 166% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 88 tok/s/user on MiniMax M3 428B (H100 FP8), M3 EAGLE delivers 428 tok/s/GPU at $0.85 per million tokens; Off delivers 171 tok/s/GPU at $2.11. M3 EAGLE is 148% cheaper per token; M3 EAGLE delivers 150% more tok/s/GPU. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution.
M3 EAGLE posts 285 tok/s/GPU for $1.25 per million tokens at 116 tok/s/user on MiniMax M3 428B (H100 FP8); Off posts 98 tok/s/GPU for $3.68. M3 EAGLE is 195% cheaper per token; M3 EAGLE delivers 191% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level. (Numbers reflect this URL's pinned 1k/1k · fp8 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | M3 EAGLE:737.4Off:277.7 | M3 EAGLE:428.3Off:171.0 | M3 EAGLE:284.8Off:97.9 |
| Cost ($/M tok) | M3 EAGLE:$0.494Off:$1.287 | M3 EAGLE:$0.853Off:$2.112 | M3 EAGLE:$1.247Off:$3.680 |
| tok/s/MW | M3 EAGLE:426244Off:160497 | M3 EAGLE:247574Off:98848 | M3 EAGLE:164599Off:56582 |
| Concurrency | M3 EAGLE:~53Off:~19 | M3 EAGLE:~21Off:~8 | M3 EAGLE:~11Off:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.