A new version of OpenAI’s Codex is powered by a new dedicated chip
This new AI coding assistant is so fast it could change how developers work.
Deep Dive
OpenAI released GPT-5.3-Codex-Spark, a lightweight version of its Codex agent designed for rapid prototyping with ultra-low latency. It's powered by Cerebras's new Wafer Scale Engine 3 chip, featuring 4 trillion transistors, as part of a massive multi-year partnership worth over $10 billion. Currently in research preview for ChatGPT Pro users, Spark focuses on real-time collaboration and fast inference, marking the first milestone in OpenAI's deep hardware integration with Cerebras.
Why It Matters
It signals a major shift towards specialized AI hardware, promising dramatically faster coding assistance and new developer workflows.