GB200 NVL72: FP4 vs FP8 Precision Comparison

Throughput at 49 tok/s/user on GLM 5/5.1 (GB200 NVL72): FP4 hits 10028 tok/s/GPU, FP8 hits 4348. Per-million costs land at $0.06 and $0.14 respectively. FP4 is 132% cheaper per token; FP4 delivers 131% more tok/s/GPU. The cost-throughput tradeoff from lower precision is only part of the picture — see the evaluation page for accuracy data.

Around the middle of the 30–109 tok/s/user interactivity band, at 69 tok/s/user on GLM 5/5.1 (GB200 NVL72): FP4 runs 8993 tok/s/GPU at $0.07/M tokens, FP8 runs 1241 at $0.49/M. FP4 is 620% cheaper per token; FP4 delivers 625% more tok/s/GPU. Precision changes affect both inference speed and model quality — consult the evaluation tab for accuracy benchmarks.

At 89 tok/s/user on GLM 5/5.1 (GB200 NVL72), FP4 delivers 5993 tok/s/GPU at $0.10 per million tokens; FP8 delivers 444 tok/s/GPU at $1.40. FP4 is 1250% cheaper per token; FP4 delivers 1249% more tok/s/GPU. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact. (Numbers reflect the default 8k/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.)