A new version of OpenAI’s Codex is powered by a new dedicated chip

OpenAI launches Codex-Spark, a new coding tool powered by Cerebras dedicated chip hardware

OpenAI has launched GPT-5.3-Codex-Spark, a lightweight coding model powered by Cerebras’ Wafer Scale Engine 3 chip. This marks the first milestone in OpenAI’s partnership with Cerebras, a specialized AI chipmaker.

Performance and efficiency improvements

The new Codex-Spark model is reported to be 15 times faster at coding tasks than its predecessor, GPT-5.3-Codex. This significant performance boost is attributed to the dedicated silicon hardware, which enables ultra-low-latency operations and improved efficiency.

Shift from general-purpose to specialized AI hardware

OpenAI has moved away from using general-purpose CPUs and instead leverages purpose-built ASICs (Application-Specific Integrated Circuits) for its AI models. This shift reflects a broader trend in AI development toward dedicated hardware for enhanced performance and reduced latency.

Industry implications

The partnership between OpenAI and Cerebras signals a strategic move to reduce reliance on Nvidia, highlighting the growing importance of specialized AI chips in accelerating AI model development and deployment.

來源：https://techcrunch.com/2026/02/12/a-new-version-of-openais-codex-is-powered-by-a-new-dedicated-chip/