---
format: "aidr-story-markdown/v1"
id: "c1132e8d8406fd97af168ce00cf221e65937285756c178617687f9309ab15abc"
canonical_url: "https://aidr.today/c1132e8d?lang=en"
title: "OpenAI Jalapeño Outperforms Nvidia GB300 With AI Code Humans Cannot Read"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-30T08:37:30.000Z"
category: "Chips"
topics: ["openai","nvidia","chips","inference"]
source_urls: ["https://x.com/firesidealpha/status/2093860831595499892","https://x.com/jukan05/status/2093873540760269133","https://x.com/Beth_Kindig/status/2093747322433700242","https://x.com/Beth_Kindig/status/2093806797349941248"]
summary: "OpenAI's Jalapeño chip delivered 1.5x to 1.9x higher tokens per watt at peak throughput and 1.7x to 3.6x lower latency than Nvidia's GB200 and GB300 on GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T models. Developed with Broadcom, the hardware utilizes AI-generated assembly and kernels written in Gluon, a low-level language built on top of Triton that OpenAI engineers cannot reason about line by line. The design approach removes human programmer usability as a primary constraint, relying instead on AI to write the compiler and optimize data movement across the hardware. OpenAI plans initial deployments before year end, with full scale operation starting in 2027."
---

# OpenAI Jalapeño Outperforms Nvidia GB300 With AI Code Humans Cannot Read

> [Open the canonical story](<https://aidr.today/c1132e8d?lang=en>)

**Published:** 2026-08-30T08:37:30.000Z
**Category:** Chips
**Topics:** openai, nvidia, chips, inference

## Summary

OpenAI's Jalapeño chip delivered 1\.5x to 1\.9x higher tokens per watt at peak throughput and 1\.7x to 3\.6x lower latency than Nvidia's GB200 and GB300 on GPT\-OSS 120B, DeepSeek R1, and Kimi K2\.5 1T models\. Developed with Broadcom, the hardware utilizes AI\-generated assembly and kernels written in Gluon, a low\-level language built on top of Triton that OpenAI engineers cannot reason about line by line\. The design approach removes human programmer usability as a primary constraint, relying instead on AI to write the compiler and optimize data movement across the hardware\. OpenAI plans initial deployments before year end, with full scale operation starting in 2027\.

## Sources

- [Story source](<https://x.com/firesidealpha/status/2093860831595499892>)
- [Supporting source](<https://x.com/jukan05/status/2093873540760269133>)
- [Story source](<https://x.com/Beth_Kindig/status/2093747322433700242>)
- [Supporting source](<https://x.com/Beth_Kindig/status/2093806797349941248>)