⭐ Key points OpenAI has released official performance data for its first self-developed inference processor, Jalapeño. According to the company, the accelerator provides performance that is 1.5 to 1.9 times per watt compared to competing systems. Tests in the InferenceX benchmark show latency times that are 1.7 to 3.6 times lower. Deployment of the chip in data centers is expected to start toward the end of 2026. OpenAI has published data [...] The post OpenAI unveiled the performance of the Jalapeño processor – 1.9 times more efficient than competitors first appeared on Technology and Entertainment Magazine - Gadgety.
Automatic translation by Google
Curated from Gadgety. Classified as Business & Technology by AI, which sometimes makes mistakes.
At TechCrunch Disrupt 2026, the AI Stage is back to dig into the single hottest topic in the community for the past few years, presented by Google for Startups.
A home's heating and cooling system can be one of the biggest drivers of utility costs.
Nvidia has reportedly agreed to acquire Hugging Face for $12.9 billion, in a deal that would give the US chipmaker control of one of the artificial intelligence industry’s most prominent platforms for sharing open-source models.
