MiniMax M3.1 Flash 428B · Chip comparison
MiniMax M3.1 Flash 428B — GB200 NVL72 vs Vera Rubin NVL72
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on MiniMax M3.1 Flash 428B. 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 376–649 tok/s/user interactivity band, at 444 tok/s/user on MiniMax M3.1 Flash 428B: GB200 NVL72 runs 14,739 tok/s/chip at $0.035 per million tokens, Vera Rubin NVL72 runs 154,491 at $0.006. Vera Rubin NVL72 is 81% cheaper per token; Vera Rubin NVL72 delivers 10× the tok/s/chip.
Setting 513 tok/s/user as the target on MiniMax M3.1 Flash 428B, GB200 NVL72 produces 9,169 tok/s/chip ($0.056 per million tokens) and Vera Rubin NVL72 produces 84,377 ($0.012). Vera Rubin NVL72 is 79% cheaper per token; Vera Rubin NVL72 delivers 9.2× the tok/s/chip.
At 581 tok/s/user interactivity on MiniMax M3.1 Flash 428B, GB200 NVL72 delivers 8,074 tok/s/chip at $0.064 per million tokens; Vera Rubin NVL72 delivers 66,931 tok/s/chip at $0.015. Vera Rubin NVL72 is 77% cheaper per token; Vera Rubin NVL72 delivers 8.3× the tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL7214,739Vera Rubin NVL72154,491 | GB200 NVL729,169Vera Rubin NVL7284,377 | GB200 NVL728,074Vera Rubin NVL7266,931 |
| Cost ($/M tok) | GB200 NVL72$0.035Vera Rubin NVL72$0.006 | GB200 NVL72$0.056Vera Rubin NVL72$0.012 | GB200 NVL72$0.064Vera Rubin NVL72$0.015 |
| tok/s/MW | GB200 NVL727,881,586Vera Rubin NVL7246,815,483 | GB200 NVL724,903,474Vera Rubin NVL7225,568,899 | GB200 NVL724,317,880Vera Rubin NVL7220,282,257 |
| Concurrency | GB200 NVL72~7Vera Rubin NVL72~204 | GB200 NVL72~3Vera Rubin NVL72~124 | GB200 NVL72~2Vera Rubin NVL72~77 |