GLM 5/5.1 · GB200 NVL72 · Precision Comparison
GB200 NVL72: FP4 vs FP8 Precision Comparison
How FP4 and FP8 precision affect GLM 5/5.1 inference on GB200 NVL72 (NVIDIA Blackwell). Throughput, latency, and cost across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
Throughput at 49 tok/s/user on GLM 5/5.1 (GB200 NVL72): FP4 hits 10,028 tok/s/chip, FP8 hits 4,348. Costs per million tokens land at $0.052 and $0.12 respectively. FP4 delivers 131% more tok/s/chip on the same chip, which makes it 57% cheaper per token. 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 8,993 tok/s/chip at $0.057 per million tokens, FP8 runs 1,241 at $0.42. FP4 delivers 7.2× the tok/s/chip on the same chip, which makes it 86% cheaper per token. 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 5,993 tok/s/chip at $0.086 per million tokens; FP8 delivers 444 tok/s/chip at $1.16. FP4 delivers 13× the tok/s/chip on the same chip, which makes it 93% cheaper per token. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact. (Numbers reflect the default 8k/1k selection for this URL — table and chart below update if you change sequence or model in the controls. Each side uses the best available serving configuration for that precision, which may include speculative decoding such as MTP where recipes exist — the same convention as the other comparison pages.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | FP410,028FP84,348 | FP48,993FP81,241 | FP45,993FP8444 |
| Cost ($/M tok) | FP4$0.052FP8$0.12 | FP4$0.057FP8$0.42 | FP4$0.086FP8$1.16 |
| tok/s/MW | FP45,362,582FP82,325,056 | FP44,808,874FP8663,610 | FP43,205,073FP8237,619 |
| Concurrency | FP4~1,404FP8~1,975 | FP4~664FP8~136 | FP4~641FP8~47 |
