DeepSeek V4.1 Flash 552B · Chip comparison
DeepSeek V4.1 Flash 552B — B300 vs Vera Rubin NVL72
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) 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 →
Near the low end of the 181–833 tok/s/user interactivity band, at 344 tok/s/user on DeepSeek V4.1 Flash 552B: B300 runs 102,075 tok/s/chip at $0.006 per million tokens, Vera Rubin NVL72 runs 2,107,103 at $0.000. Vera Rubin NVL72 is 92% cheaper per token; Vera Rubin NVL72 delivers 21× the tok/s/chip.
Setting 507 tok/s/user as the target on DeepSeek V4.1 Flash 552B, B300 produces 58,107 tok/s/chip ($0.011 per million tokens) and Vera Rubin NVL72 produces 1,296,940 ($0.001). Vera Rubin NVL72 is 93% cheaper per token; Vera Rubin NVL72 delivers 22× the tok/s/chip.
At 670 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, B300 delivers 24,640 tok/s/chip at $0.025 per million tokens; Vera Rubin NVL72 delivers 648,662 tok/s/chip at $0.002. Vera Rubin NVL72 is 94% cheaper per token; Vera Rubin NVL72 delivers 26× the tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300102,075Vera Rubin NVL722,107,103 | B30058,107Vera Rubin NVL721,296,940 | B30024,640Vera Rubin NVL72648,662 |
| Cost ($/M tok) | B300$0.006Vera Rubin NVL72$0.000 | B300$0.011Vera Rubin NVL72$0.001 | B300$0.025Vera Rubin NVL72$0.002 |
| tok/s/MW | B30053,723,528Vera Rubin NVL72638,516,027 | B30030,582,615Vera Rubin NVL72393,012,244 | B30012,968,628Vera Rubin NVL72196,564,274 |
| Concurrency | B300~45Vera Rubin NVL72~11,236 | B300~28Vera Rubin NVL72~8,441 | B300~13Vera Rubin NVL72~3,414 |