gpt-oss 120B · Chip comparison
gpt-oss 120B — H100 vs MI300X
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and MI300X (AMD CDNA 3) on gpt-oss 120B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Setting 89 tok/s/user as the target on gpt-oss 120B, H100 produces 6,413 tok/s/chip ($0.051 per million tokens) and MI300X produces 5,144 ($0.051). H100 is 1% cheaper per token; H100 delivers 25% more tok/s/chip.
At 142 tok/s/user interactivity on gpt-oss 120B, H100 delivers 4,180 tok/s/chip at $0.078 per million tokens; MI300X delivers 3,406 tok/s/chip at $0.077. Cost per token is essentially tied; H100 delivers 23% more tok/s/chip at this point.
H100 posts 2,714 tok/s/chip for $0.12 per million tokens at 194 tok/s/user on gpt-oss 120B; MI300X posts 1,102 tok/s/chip for $0.24. H100 is 50% cheaper per token; H100 delivers 146% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H1006,413MI300X5,144 | H1004,180MI300X3,406 | H1002,714MI300X1,102 |
| Cost ($/M tok) | H100$0.051MI300X$0.051 | H100$0.078MI300X$0.077 | H100$0.12MI300X$0.24 |
| tok/s/MW | H1004,680,812MI300X3,700,621 | H1003,051,141MI300X2,450,139 | H1001,981,084MI300X792,671 |
| Concurrency | H100~32MI300X~13 | H100~14MI300X~5 | H100~6MI300X~7 |