B200: FP4 vs BF16 Precision Comparison

FP4 posts 4731 tok/s/GPU for $0.11 per million tokens at 112 tok/s/user on Qwen 3.5 397B-A17B (B200); BF16 posts 1216 tok/s/GPU for $0.46. FP4 is 306% cheaper per token; FP4 delivers 289% more tok/s/GPU. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 166 tok/s/user on Qwen 3.5 397B-A17B (B200): FP4 hits 1864 tok/s/GPU, BF16 hits 660. Per-million costs land at $0.29 and $0.82 respectively. FP4 is 181% cheaper per token; FP4 delivers 182% more tok/s/GPU. The cost-throughput tradeoff from lower precision is only part of the picture — see the evaluation page for accuracy data.

Toward the upper edge of the 59–272 tok/s/user interactivity band, at 219 tok/s/user on Qwen 3.5 397B-A17B (B200): FP4 runs 1014 tok/s/GPU at $0.53/M tokens, BF16 runs 299 at $1.88/M. FP4 is 254% cheaper per token; FP4 delivers 239% more tok/s/GPU. Precision changes affect both inference speed and model quality — consult the evaluation tab for accuracy benchmarks. (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.)