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 — 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 →
At 226 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, B300 delivers 18355 tok/s/chip at $0.03 per million tokens; H100 delivers 2507 tok/s/chip at $0.13. B300 is 279% cheaper per token; B300 delivers 632% more tok/s/chip at this point.
B300 posts 16338 tok/s/chip for $0.04 per million tokens at 243 tok/s/user on DeepSeek V4.1 Flash 552B; H100 posts 2481 tok/s/chip for $0.13. B300 is 241% cheaper per token; B300 delivers 558% more tok/s/chip.
Throughput at 259 tok/s/user on DeepSeek V4.1 Flash 552B: B300 hits 14450 tok/s/chip, H100 hits 2466. Per-million costs land at $0.04 and $0.13 respectively. B300 is 203% cheaper per token; B300 delivers 486% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:18354.5H100:2507.3 | B300:16338.1H100:2481.2 | B300:14450.3H100:2465.9 |
| Cost ($/M tok) | B300:$0.034H100:$0.130 | B300:$0.038H100:$0.131 | B300:$0.043H100:$0.132 |
| tok/s/MW | B300:9660282H100:1830166 | B300:8599009H100:1811072 | B300:7605438H100:1799936 |
| Concurrency | B300:~10H100:~2 | B300:~9H100:~1 | B300:~8H100:~1 |
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.
