Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) 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.
At 83 tok/s/user interactivity on gpt-oss 120B, B200 delivers 37880 tok/s/chip at $0.01 per million tokens; MI300X delivers 5465 tok/s/chip at $0.05. B200 is 281% cheaper per token; B200 delivers 593% more tok/s/chip at this point.
B200 posts 25790 tok/s/chip for $0.02 per million tokens at 138 tok/s/user on gpt-oss 120B; MI300X posts 3491 tok/s/chip for $0.08. B200 is 306% cheaper per token; B200 delivers 639% more tok/s/chip.
Throughput at 192 tok/s/user on gpt-oss 120B: B200 hits 17979 tok/s/chip, MI300X hits 1133. Per-million costs land at $0.03 and $0.23 respectively. B200 is 771% cheaper per token; B200 delivers 1487% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:37879.8MI300X:5465.2 | B200:25789.8MI300X:3491.5 | B200:17978.6MI300X:1132.9 |
| Cost ($/M tok) | B200:$0.013MI300X:$0.048 | B200:$0.019MI300X:$0.076 | B200:$0.027MI300X:$0.233 |
| tok/s/MW | B200:22151917MI300X:3931798 | B200:15081733MI300X:2511856 | B200:10513787MI300X:815020 |
| Concurrency | B200:~49MI300X:~15 | B200:~32MI300X:~5 | B200:~20MI300X:~7 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr: VR200: 3.61H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68Jalapeño (Teacup): 1.47
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
gpt-oss 120B 8K / 1K Total Tokens per $1 TCO vs. Interactivity