DeepSeek V4.1 Flash 552B · Chip comparison

DeepSeek V4.1 Flash 552B — B200 vs Vera Rubin NVL72

Head-to-head AI inference benchmark comparison of B200 (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 →

B200 / Vera Rubin NVL72 on DeepSeek V4.1 Flash 552B at 310 tok/s/user: 69,185 / 2,334,036 tok/s/chip, $0.007 / $0.000 per million tokens. Vera Rubin NVL72 is 94% cheaper per token; Vera Rubin NVL72 delivers 34× the tok/s/chip.

Around the middle of the 181–699 tok/s/user interactivity band, at 440 tok/s/user on DeepSeek V4.1 Flash 552B: B200 runs 29,027 tok/s/chip at $0.017 per million tokens, Vera Rubin NVL72 runs 1,560,612 at $0.001. Vera Rubin NVL72 is 96% cheaper per token; Vera Rubin NVL72 delivers 54× the tok/s/chip.

Setting 570 tok/s/user as the target on DeepSeek V4.1 Flash 552B, B200 produces 10,537 tok/s/chip ($0.046 per million tokens) and Vera Rubin NVL72 produces 1,066,719 ($0.001). Vera Rubin NVL72 is 98% cheaper per token; Vera Rubin NVL72 delivers 101× 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.)

View performance per dollar →

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)
B20069,185Vera Rubin NVL722,334,036
B20029,027Vera Rubin NVL721,560,612
B20010,537Vera Rubin NVL721,066,719
Cost ($/M tok)
B200$0.007Vera Rubin NVL72$0.000
B200$0.017Vera Rubin NVL72$0.001
B200$0.046Vera Rubin NVL72$0.001
tok/s/MW
B20040,459,338Vera Rubin NVL72707,283,605
B20016,974,824Vera Rubin NVL72472,912,651
B2006,162,128Vera Rubin NVL72323,248,178
Concurrency
B200~37Vera Rubin NVL72~11,297
B200~16Vera Rubin NVL72~10,580
B200~4Vera Rubin NVL72~6,376