Export
The link always opens the latest version of this page.
Ask an AI about it
Open in ChatGPTA new chat that reads this entry from the site Open in ClaudeThe same, in ClaudeShare the link instead of a file, and they will always see the latest version.
Files are a snapshot of today. The live page has anything newer.
OpenAI Jalapeño
OpenAI's custom inference chip for LLM inference, announced at Hot Chips, developed in partnership with Broadcom in ~16 months, outperforms Nvidia Blackwell on perf/W metrics.
- Category
- inference accelerator 1 source
A custom inference accelerator chip developed by openai for large language model inference operations.
Jalapeño is built on HBM4 memory technology 1 and is purpose-designed for inference workloads 2. The chip was developed in partnership with Broadcom over approximately 16 months and was announced at Hot Chips. According to performance metrics presented at the announcement, Jalapeño achieves superior performance-per-watt efficiency compared to Nvidia's Blackwell architecture.
Facts 2
| Attribute | Value | Confidence | Sources | Cited |
|---|---|---|---|---|
| Vendor | OpenAI | low | 1 | 1 |
| Category | inference accelerator | low | 1 | 1 |
Platform 2
Neighbourhood
Developed by 2
Involved in 1
Official links
Frequently mentioned together 15
Sources 9
- OpenAI Jalapeño: Better than Nvidia Blackwell · Hacker News (YC)
- OpenAI's upcoming Jalapeño chip looks like it'll be an inference beast · The Register: Security
- Sovereign AI, same landlord — thoughts on the physical AI race · The Robot Report
- Tokens Per Watt: Why Your Context Window Is a Power Decision · HackerNoon
- The U.S.-China AI Race is Changing the Stakes for Taiwan · Council on Foreign Relations
5 of 9 sources shown · Pro unlocks the full list