OpenAI's new chip has Indian and Japanese food inside it
- OpenAI’s new inference chip, Jalapeño, has shown higher performance per watt than Nvidia’s Blackwell in SemiAnalysis benchmark tests.
- The chip’s development took about 16 months, highlighting the rapid pace of OpenAI’s ASIC development.
- Jalapeño is designed as a generalised AI inference chip capable of running different models and workloads, rather than being limited to OpenAI models.
- The use of HBM4 gives Jalapeño specifications comparable with flagship GPUs from Nvidia and AMD.
Source: newsletter.semianalysis.com
Get 100 stories a day on the app
Read, watch & listen · Free on iOS & Android
