OpenAI has released GPT-5.3-Codex-Spark, a lightweight version of its Codex tool, focusing on speed and efficiency for developers.
OpenAI has unveiled a streamlined iteration of its programming assistant, Codex, named GPT-5.3-Codex-Spark.
Now in research preview, it enables developers to create projects more swiftly. pic.twitter.com/85LzDOgcQj
— OpenAI (@OpenAI) February 12, 2026
Targeted for speed, this version leverages cutting-edge technology from Cerebras, featuring their Wafer Scale Engine 3, a chip composed of four trillion transistors.
Previously announced in January, the collaboration between OpenAI and Cerebras aims to enhance AI response rates significantly.
OpenAI remarked, “Integrating Cerebras into our computing solutions aims to significantly speed up AI response times.”
The Spark chip is noted as a pivotal advancement in their partnership, facilitating swift, real-time operations essential for efficient programming tasks.
Unlike the original GPT-5.3 Codex, which is geared toward intricate projects, Codex-Spark is positioned as a "daily productivity driver" aiding in quick prototyping.
OpenAI stated that Codex-Spark is aimed at reducing latency and fostering real-time collaboration, catering to immediate development needs and longer-term projects requiring extensive reasoning.
They described it as a foundational step towards a dual-functioning Codex model, focusing on both interactive and complex tasks.
This launch follows the release of a standalone Codex application by OpenAI in February 2026, marking a significant evolution in AI-assisted programming tools.
