DeepSeek V4 Pro 1.6T — H200 vs MI325X Performance per Dollar
Cost per million tokens of H200 (NVIDIA Hopper) versus MI325X (AMD CDNA 3) on DeepSeek V4 Pro 1.6T. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
At 13 tok/s/user on DeepSeek V4 Pro 1.6T, H200 costs $0.92 per million tokens; MI325X costs $0.85. MI325X is 8% more cost-efficient at this operating point.
H200 edges MI325X at 20 tok/s/user on DeepSeek V4 Pro 1.6T — $0.99 per million tokens versus $1.75, a 77% cost-per-token gap.
Push DeepSeek V4 Pro 1.6T to 27 tok/s/user and H200 lands at $1.12 per million tokens against MI325X's $3.23 — H200 pulls ahead by 189%. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Chip pricing (owning hyperscaler): H200 $1.22/chip/hr · MI325X $1.10/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | H200:$0.917MI325X:$0.852 | H200:$0.991MI325X:$1.754 | H200:$1.119MI325X:$3.235 |
| Concurrency | H200:~117MI325X:~113 | H200:~75MI325X:~37 | H200:~48MI325X:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.