Kimi K2.5/K2.6/K2.7-Code 1T — GB200 NVL72 vs Jalapeño
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and Jalapeño (OpenAI Jalapeño) on Kimi K2.5/K2.6/K2.7-Code 1T. 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.
Near the low end of the 61–180 tok/s/user interactivity band, at 90 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: GB200 NVL72 runs 1904 tok/s/chip at $0.27/M tokens, Jalapeño runs 11469 at $0.03/M. Jalapeño is 782% cheaper per token; Jalapeño delivers 502% more tok/s/chip.
Setting 120 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, GB200 NVL72 produces 1285 tok/s/chip ($0.40 per million tokens) and Jalapeño produces 9816 ($0.04). Jalapeño is 1018% cheaper per token; Jalapeño delivers 664% more tok/s/chip.
At 150 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, GB200 NVL72 delivers 558 tok/s/chip at $0.93 per million tokens; Jalapeño delivers 7858 tok/s/chip at $0.04. Jalapeño is 1963% cheaper per token; Jalapeño delivers 1309% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:1904.1Jalapeño:11468.9 | GB200 NVL72:1285.5Jalapeño:9815.7 | GB200 NVL72:557.8Jalapeño:7857.7 |
| Cost ($/M tok) | GB200 NVL72:$0.271Jalapeño:$0.031 | GB200 NVL72:$0.402Jalapeño:$0.036 | GB200 NVL72:$0.926Jalapeño:$0.045 |
| tok/s/MW | GB200 NVL72:1018217Jalapeño:10194612 | GB200 NVL72:687422Jalapeño:8725091 | GB200 NVL72:298299Jalapeño:6984582 |
| Concurrency | GB200 NVL72:~151Jalapeño:~1087 | GB200 NVL72:~39Jalapeño:~595 | GB200 NVL72:~23Jalapeño:~324 |