MiniMax M3 428B · MI355X · Precision comparison

MI355X: FP4 vs FP8 precision comparison

How FP4 and FP8 precision affect MiniMax M3 428B inference on MI355X (AMD CDNA 4). Throughput, latency, and cost across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.

At 82 tok/s/user on MiniMax M3 428B (MI355X), FP4 delivers 6,206 tok/s/chip at $0.067 per million tokens; FP8 delivers 4,257 tok/s/chip at $0.098. FP4 delivers 46% more tok/s/chip on the same chip, which makes it 31% cheaper per token. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP4 posts 3,868 tok/s/chip for $0.11 per million tokens at 157 tok/s/user on MiniMax M3 428B (MI355X); FP8 posts 2,254 tok/s/chip for $0.18. FP4 delivers 72% more tok/s/chip on the same chip, which makes it 42% cheaper per token. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 232 tok/s/user on MiniMax M3 428B (MI355X): FP4 hits 2,229 tok/s/chip, FP8 hits 1,259. Costs per million tokens land at $0.19 and $0.33 respectively. FP4 delivers 77% more tok/s/chip on the same chip, which makes it 43% cheaper per token. The cost-throughput tradeoff from lower precision is only part of the picture — see the evaluation page for accuracy data. (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.)

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.

Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/chip)
FP46,206FP84,257
FP43,868FP82,254
FP42,229FP81,259
Cost ($/M tok)
FP4$0.067FP8$0.098
FP4$0.11FP8$0.18
FP4$0.19FP8$0.33
tok/s/MW
FP42,969,329FP82,036,667
FP41,850,770FP81,078,654
FP41,066,423FP8602,564
Concurrency
FP4~40FP8~28
FP4~13FP8~8
FP4~5FP8~3
MiniMax M3 428B: FP4 versus FP8 throughput and cost at matched interactivity levels on MI355X
FP4 versus FP8 throughput and cost per million tokens on MI355X for this comparison's canonical default workload
Open full-size image ↗