OpenAI Releases GPT-5.3-Codex-Spark Powered by Cerebras Chip

OpenAI has released GPT-5.3-Codex-Spark, a lightweight version of its Codex model designed for faster inference and real-time software development. The model is powered by a dedicated chip from Cerebras Systems, marking the first major integration between the two companies following their recently announced multi-year partnership.

Codex-Spark runs on Cerebras’ Wafer Scale Engine 3, a third-generation AI processor with 4 trillion transistors. The hardware enables the model to process over 1,000 tokens per second, allowing near-instant feedback during live coding. OpenAI describes Spark as optimized for rapid iteration and collaborative work, complementing the original Codex model designed for longer, more complex tasks.

The new Codex-Spark model is available as a research preview for ChatGPT Pro users through the Codex app, command-line interface, and VS Code extension. API access is being extended to selected design partners. OpenAI stated that this release is the first step toward offering dual operational modes in Codex: one for real-time, low-latency collaboration and another for extended reasoning and execution.

Cerebras’ involvement in the project builds on its growing role in large-scale AI infrastructure. The company recently secured $1 billion in new funding and continues to expand its wafer-scale computing systems to support high-performance AI workloads for partners like OpenAI.