B300: FP8 vs BF16 Precision Comparison

At 97 tok/s/user on Qwen 3.5 397B-A17B (B300), FP8 delivers 1815 tok/s/GPU at $0.36 per million tokens; BF16 delivers 1475 tok/s/GPU at $0.44. FP8 is 22% cheaper per token; FP8 delivers 23% more tok/s/GPU. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP8 posts 977 tok/s/GPU for $0.66 per million tokens at 150 tok/s/user on Qwen 3.5 397B-A17B (B300); BF16 posts 774 tok/s/GPU for $0.83. FP8 is 26% cheaper per token; FP8 delivers 26% more tok/s/GPU. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 202 tok/s/user on Qwen 3.5 397B-A17B (B300): FP8 hits 662 tok/s/GPU, BF16 hits 516. Per-million costs land at $0.98 and $1.22 respectively. FP8 is 25% cheaper per token; FP8 delivers 28% 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.)