OpenAI's upcoming Jalapeño chip aims for high inference performance
UKPulse News Desk
OpenAI is developing a new chip, codenamed Jalapeño, which is reportedly designed for inference tasks and could offer significant performance advantages.
- OpenAI's upcoming Jalapeño chip is expected to be an 'inference beast'.
- The chip is projected to feature 128 chips, 1.7 exaFLOPS, and 27 TB of HBM.
- Jalapeño could provide OpenAI with an advantage over Nvidia's Blackwell and Rubin architectures.
OpenAI is reportedly developing a new chip, codenamed Jalapeño, which is anticipated to be a powerful component for inference tasks. This upcoming chip is expected to feature 128 individual chips, deliver 1.7 exaFLOPS of performance, and include 27 terabytes of High Bandwidth Memory (HBM).
These specifications suggest that the Jalapeño chip could give OpenAI a competitive edge over existing architectures, specifically mentioning Nvidia's Blackwell and potentially Rubin.
Why this matters: The development of custom AI chips by companies like OpenAI could influence the landscape of AI hardware and the capabilities of future AI models.