OpenAI's Jalapeño Chip Beats Nvidia GB200/GB300 in Efficiency

Beth_Kindig · x · 2026-08-30

OpenAI revealed that its in-house Jalapeño chip delivered 1.5X-1.9X higher tokens per watt and 1.7X-3.6X lower latency compared to Nvidia's GB200 and GB300 across GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T models. The company is preparing for scale operations and initial deployments before year-end.

Related event: OpenAI's In-House Jalapeño Chip Reportedly Beats Nvidia on Efficiency(3 posts)→

Original post →

More from Infra

Infra channel →