Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on MiniMax M3 428B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics β 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 β
At 178 tok/s/user on MiniMax M3 428B, B200 delivers 33359 tok/s/chip at $0.01 per million tokens; MI325X hasn't been benchmarked at this target.
B200 hits 28503 tok/s/chip for $0.02 per million tokens at 291 tok/s/user on MiniMax M3 428B. No MI325X data at this operating point.
B200: 16650 tok/s/chip, $0.03 per million tokens at 405 tok/s/user on MiniMax M3 428B. MI325X is unmeasured here. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:33358.8MI325X:β | B200:28502.5MI325X:β | B200:16649.7MI325X:β |
| Cost ($/M tok) | B200:$0.014MI325X:β | B200:$0.017MI325X:β | B200:$0.029MI325X:β |
| tok/s/MW | B200:19508070MI325X:β | B200:16668142MI325X:β | B200:9736694MI325X:β |
| Concurrency | B200:~20MI325X:β | B200:~20MI325X:β | B200:~11MI325X:β |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
