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

On Thursday, OpenAI announced the release of a lightweight version of its Codex coding tool, the latest version of which. OpenAI was launched at the beginning of this month. GPT-5.3-Codex-Spark is described by the company as a “small version” of that model, which is designed to be faster. To achieve this, OpenAI has brought a dedicated chip from its hardware partner Cerebras, which represents a new phase of integration in the company’s architecture.

The agreement between Cerebras and OpenAI was announced last month, where OpenAI he said that reached a multi-year contract with the company worth more than $10 billion. “Incorporating Cerebras into our mix of compute solutions is making our AI more responsive,” the company said at the time. Now, OpenAI calls Spark “the first priority” in this relationship.

Spark, which OpenAI says is designed for high-speed collaboration, virtual reality and “rapid iteration,” will be powered by Cerebras’ Wafer Scale Engine 3. WSE-3 and Cerebras’ the third generation waferscale megachip, adorned with 4 trillion transistors. OpenAI describes the new lightweight tool as “a driver for everyday productivity, helping users with rapid prototyping” instead of the long, heavy workloads that the original 5.3 was designed for. Spark is currently enjoying a preview of the ChatGPT Pro user survey in the Codex app.

In a tweet ahead of the announcement, CEO Sam Altman appeared to be hinting at the new model. “We have a special feature we’re launching for Codex users on the Pro plan later today,” Altman wrote. “It brings me joy.”

In its official statement, OpenAI emphasized Spark as being designed to be very low-level on Codex. “Codex-Spark is the first step towards a Codex that works in two complementary ways: real collaboration when you need to iterate quickly, and long-term projects when you need to think deeply and execute,” OpenAI shared. The company added that Cerebras chips excel in supporting “workflows that require very low latency.”

Macerebras has been there for more than ten years but, in the era of AI, it has played a prominent role in the technology industry. Last week, the company he announced that it raised $1 billion in new capital $23 billion. The company has done it has already been announced its plans to pursue an IPO.

“What we’re really excited about for GPT-5.3-Codex-Spark is working with OpenAI and the developer community to find out what makes it possible – new ways to connect, new ways to use it, and a different interface,” Sean Lie, CTO and co-founder of Cerebras, said in a statement. “This show is just the beginning.”