B300: FP4 vs FP8 Precision Comparison

At 111 tok/s/user on MiniMax M3 428B (B300), FP4 delivers 2245 tok/s/GPU at $0.29 per million tokens; FP8 delivers 1155 tok/s/GPU at $0.56. FP4 is 94% cheaper per token; FP4 delivers 94% more tok/s/GPU. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP4 posts 1029 tok/s/GPU for $0.64 per million tokens at 212 tok/s/user on MiniMax M3 428B (B300); FP8 posts 539 tok/s/GPU for $1.19. FP4 is 85% cheaper per token; FP4 delivers 91% more tok/s/GPU. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 312 tok/s/user on MiniMax M3 428B (B300): FP4 hits 347 tok/s/GPU, FP8 hits 259. Per-million costs land at $1.91 and $2.53 respectively. FP4 is 32% cheaper per token; FP4 delivers 34% more tok/s/GPU. The cost-throughput tradeoff from lower precision is only part of the picture — see the evaluation page for accuracy data. (Numbers reflect the default 1k/1k selection for this URL — table and chart below update if you change sequence or model in the controls. Each side uses the best available serving configuration for that precision, which may include speculative decoding such as MTP where recipes exist — the same convention as the other comparison pages.)