MI355X: FP4 vs BF16 Precision Comparison

At 64 tok/s/user on Qwen 3.5 397B-A17B (MI355X), FP4 delivers 2220 tok/s/chip at $0.19 per million tokens; BF16 delivers 879 tok/s/chip at $0.47. FP4 is 153% cheaper per token; FP4 delivers 153% more tok/s/chip. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP4 posts 1475 tok/s/chip for $0.28 per million tokens at 99 tok/s/user on Qwen 3.5 397B-A17B (MI355X); BF16 posts 597 tok/s/chip for $0.70. FP4 is 147% cheaper per token; FP4 delivers 147% more tok/s/chip. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 133 tok/s/user on Qwen 3.5 397B-A17B (MI355X): FP4 hits 1104 tok/s/chip, BF16 hits 279. Per-million costs land at $0.38 and $1.49 respectively. FP4 is 295% cheaper per token; FP4 delivers 295% more tok/s/chip. 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.)