OpenAI has officially launched GPT-5.3-Codex-Spark, a new lightweight iteration of its agentic coding tool, Codex, which was initially released earlier this month. This new model, described by OpenAI as a “smaller version” of the larger 5.3 model, is specifically engineered for significantly faster inference and rapid iteration in coding workflows.

A critical component enabling Spark’s high-speed inference is the dedicated chip supplied by OpenAI’s hardware partner, Cerebras. This collaboration signifies a deeper level of integration within OpenAI’s physical infrastructure. The partnership between Cerebras and OpenAI was initially announced last month, encompassing a multi-year agreement valued at over $10 billion. OpenAI stated that “Integrating Cerebras into our mix of compute solutions is all about making our AI respond much faster,” and now refers to Spark as the “first milestone” in this significant relationship.

GPT-5.3-Codex-Spark is powered by Cerebras’ advanced Wafer Scale Engine 3 (WSE-3), which is the company’s third-generation waferscale megachip, boasting an impressive 4 trillion transistors. OpenAI positions this new lightweight tool as a “daily productivity driver” intended to assist users with rapid prototyping and real-time collaboration, distinguishing it from the longer, heavier computational tasks that the original 5.3 model is designed to handle.

Currently, Spark is available in a research preview for ChatGPT Pro users through the Codex app. Ahead of the official announcement, OpenAI CEO Sam Altman subtly hinted at the launch via a tweet, stating, “We have a special thing launching to Codex users on the Pro plan later today. It sparks joy for me.” OpenAI’s official statement further emphasized Spark’s design for achieving the lowest possible latency on Codex, envisioning a future Codex that operates in two complementary modes: one for real-time collaboration and rapid iteration (Spark), and another for long-running tasks requiring deeper reasoning and execution.

Cerebras, a company with over a decade of experience, has seen its prominence grow significantly in the AI era. Just last week, it announced a successful fundraising round, securing $1 billion in fresh capital and achieving a valuation of $23 billion, with intentions to pursue an IPO. Sean Lie, CTO and co-founder of Cerebras, expressed his excitement, noting, “What excites us most about GPT-5.3-Codex-Spark is partnering with OpenAI and the developer community to discover what fast inference makes possible — new interaction patterns, new use cases, and a fundamentally different model experience. This preview is just the beginning.”