DeepSeek V4.1 Flash 552B · Chip comparison
DeepSeek V4.1 Flash 552B — H100 vs Vera Rubin NVL72
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and Vera Rubin NVL72 (NVIDIA Vera Rubin) 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 →
Throughput at 217 tok/s/user on DeepSeek V4.1 Flash 552B: H100 hits 5,569 tok/s/chip, Vera Rubin NVL72 hits 2,835,965. Costs per million tokens land at $0.058 and $0.000 respectively. Vera Rubin NVL72 is 99% cheaper per token; Vera Rubin NVL72 delivers 509× the tok/s/chip.
H100 / Vera Rubin NVL72 on DeepSeek V4.1 Flash 552B at 252 tok/s/user: 4,305 / 2,671,988 tok/s/chip, $0.075 / $0.000 per million tokens. Vera Rubin NVL72 is 99.5% cheaper per token; Vera Rubin NVL72 delivers 621× the tok/s/chip.
Toward the upper edge of the 181–323 tok/s/user interactivity band, at 288 tok/s/user on DeepSeek V4.1 Flash 552B: H100 runs 2,910 tok/s/chip at $0.11 per million tokens, Vera Rubin NVL72 runs 2,473,533 at $0.000. Vera Rubin NVL72 is 99.6% cheaper per token; Vera Rubin NVL72 delivers 850× the tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H1005,569Vera Rubin NVL722,835,965 | H1004,305Vera Rubin NVL722,671,988 | H1002,910Vera Rubin NVL722,473,533 |
| Cost ($/M tok) | H100$0.058Vera Rubin NVL72$0.000 | H100$0.075Vera Rubin NVL72$0.000 | H100$0.11Vera Rubin NVL72$0.000 |
| tok/s/MW | H1004,064,621Vera Rubin NVL72859,383,367 | H1003,142,395Vera Rubin NVL72809,693,203 | H1002,123,826Vera Rubin NVL72749,555,572 |
| Concurrency | H100~6Vera Rubin NVL72~13,145 | H100~4Vera Rubin NVL72~11,604 | H100~2Vera Rubin NVL72~11,359 |