DeepSeek R1 — GB200 NVL72 vs Jalapeño
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and Jalapeño (OpenAI Jalapeño) on DeepSeek R1. 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.
GB200 NVL72 / Jalapeño on DeepSeek R1 at 124 tok/s/user: 8957 / 11215 tok/s/chip, $0.06 / $0.03 per million tokens. Jalapeño is 83% cheaper per token; Jalapeño delivers 25% more tok/s/chip.
Around the middle of the 66–299 tok/s/user interactivity band, at 183 tok/s/user on DeepSeek R1: GB200 NVL72 runs 2358 tok/s/chip at $0.22/M tokens, Jalapeño runs 7781 at $0.05/M. Jalapeño is 383% cheaper per token; Jalapeño delivers 230% more tok/s/chip.
Setting 241 tok/s/user as the target on DeepSeek R1, GB200 NVL72 produces 848 tok/s/chip ($0.61 per million tokens) and Jalapeño produces 2606 ($0.14). Jalapeño is 350% cheaper per token; Jalapeño delivers 207% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:8956.7Jalapeño:11215.4 | GB200 NVL72:2357.8Jalapeño:7780.5 | GB200 NVL72:848.2Jalapeño:2605.8 |
| Cost ($/M tok) | GB200 NVL72:$0.058Jalapeño:$0.031 | GB200 NVL72:$0.219Jalapeño:$0.045 | GB200 NVL72:$0.609Jalapeño:$0.135 |
| tok/s/MW | GB200 NVL72:4789664Jalapeño:9969252 | GB200 NVL72:1260838Jalapeño:6916013 | GB200 NVL72:453607Jalapeño:2316229 |
| Concurrency | GB200 NVL72:~679Jalapeño:~685 | GB200 NVL72:~80Jalapeño:~255 | GB200 NVL72:~15Jalapeño:~48 |