Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on MiniMax M3.1 Flash 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
Setting 84 tok/s/user as the target on MiniMax M3.1 Flash 428B, B300 produces 62822 tok/s/chip ($0.01 per million tokens) and H100 produces 5732 ($0.06). B300 is 467% cheaper per token; B300 delivers 996% more tok/s/chip.
At 152 tok/s/user interactivity on MiniMax M3.1 Flash 428B, B300 delivers 60952 tok/s/chip at $0.01 per million tokens; H100 delivers 4486 tok/s/chip at $0.07. B300 is 603% cheaper per token; B300 delivers 1259% more tok/s/chip at this point.
B300 posts 39745 tok/s/chip for $0.02 per million tokens at 220 tok/s/user on MiniMax M3.1 Flash 428B; H100 posts 3595 tok/s/chip for $0.09. B300 is 472% cheaper per token; B300 delivers 1006% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:62822.1H100:5732.4 | B300:60952.0H100:4486.1 | B300:39744.7H100:3594.6 |
| Cost ($/M tok) | B300:$0.010H100:$0.057 | B300:$0.010H100:$0.072 | B300:$0.016H100:$0.090 |
| tok/s/MW | B300:33064267H100:4184237 | B300:32080018H100:3274527 | B300:20918240H100:2623785 |
| Concurrency | B300:~53H100:~9 | B300:~38H100:~7 | B300:~21H100:~3 |
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.
