Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on DeepSeek V4.1 Flash 552B. 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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Setting 71 tok/s/user as the target on DeepSeek V4.1 Flash 552B, B300 produces 129751 tok/s/chip ($0.00 per million tokens) and MI325X produces 7116 ($0.04). B300 is 787% cheaper per token; B300 delivers 1723% more tok/s/chip.
At 116 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, B300 delivers 68455 tok/s/chip at $0.01 per million tokens; MI325X delivers 4858 tok/s/chip at $0.06. B300 is 586% cheaper per token; B300 delivers 1309% more tok/s/chip at this point.
B300 posts 54097 tok/s/chip for $0.01 per million tokens at 160 tok/s/user on DeepSeek V4.1 Flash 552B; MI325X posts 3014 tok/s/chip for $0.10. B300 is 774% cheaper per token; B300 delivers 1695% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:129751.4MI325X:7116.1 | B300:68454.8MI325X:4857.8 | B300:54096.7MI325X:3014.3 |
| Cost ($/M tok) | B300:$0.005MI325X:$0.043 | B300:$0.009MI325X:$0.063 | B300:$0.012MI325X:$0.101 |
| tok/s/MW | B300:68290218MI325X:4210722 | B300:36028840MI325X:2874425 | B300:28471946MI325X:1783617 |
| Concurrency | B300:~40MI325X:~11 | B300:~32MI325X:~6 | B300:~15MI325X:~4 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
