B200: FP4 vs FP8 Precision Comparison

Near the low end of the 15–389 tok/s/user interactivity band, at 108 tok/s/user on MiniMax M3 428B (B200): FP4 runs 2058 tok/s/GPU at $0.26/M tokens, FP8 runs 1429 at $0.40/M. FP4 is 52% cheaper per token; FP4 delivers 44% more tok/s/GPU. Precision changes affect both inference speed and model quality — consult the evaluation tab for accuracy benchmarks.

At 202 tok/s/user on MiniMax M3 428B (B200), FP4 delivers 834 tok/s/GPU at $0.65 per million tokens; FP8 delivers 542 tok/s/GPU at $1.00. FP4 is 53% cheaper per token; FP4 delivers 54% more tok/s/GPU. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP4 posts 286 tok/s/GPU for $1.89 per million tokens at 296 tok/s/user on MiniMax M3 428B (B200); FP8 posts 190 tok/s/GPU for $2.93. FP4 is 55% cheaper per token; FP4 delivers 51% more tok/s/GPU. Quantization-level accuracy differences are tracked on the evaluation tab. (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.)