Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on DeepSeek V4.1 Flash 552B. 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 119 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, B300 delivers 261322 tok/s/chip at $0.00 per million tokens; H100 delivers 15331 tok/s/chip at $0.02. B300 is 782% cheaper per token; B300 delivers 1604% more tok/s/chip at this point.
B300 posts 132589 tok/s/chip for $0.00 per million tokens at 179 tok/s/user on DeepSeek V4.1 Flash 552B; H100 posts 7541 tok/s/chip for $0.04. B300 is 810% cheaper per token; B300 delivers 1658% more tok/s/chip.
Throughput at 238 tok/s/user on DeepSeek V4.1 Flash 552B: B300 hits 78172 tok/s/chip, H100 hits 4276. Per-million costs land at $0.01 and $0.08 respectively. B300 is 847% cheaper per token; B300 delivers 1728% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:261322.1H100:15331.5 | B300:132589.1H100:7541.3 | B300:78172.4H100:4275.5 |
| Cost ($/M tok) | B300:$0.002H100:$0.021 | B300:$0.005H100:$0.043 | B300:$0.008H100:$0.076 |
| tok/s/MW | B300:137537961H100:11190863 | B300:69783751H100:5504575 | B300:41143387H100:3120818 |
| Concurrency | B300:~132H100:~18 | B300:~37H100:~8 | B300:~24H100:~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.
