On Thursday, OpenAI introduced the discharge of a lightweight model of its agentic coding software Codex, the newest mannequin of which OpenAI launched earlier this month. GPT-5.3-Codex-Spark is described by the corporate as a “smaller model” of that mannequin, one that’s designed for sooner inference. To energy that inference, OpenAI has introduced in a devoted chip from its {hardware} associate Cerebras, marking a brand new degree of integration within the firm’s bodily infrastructure.
The partnership between Cerebras and OpenAI was introduced final month, when OpenAI mentioned that it had reached a multi-year settlement with the agency value over $10 billion. “Integrating Cerebras into our mixture of compute options is all about making our AI reply a lot sooner,” the corporate mentioned on the time. Now, OpenAI calls Spark the “first milestone” in that relationship.
Spark, which OpenAI says is designed for swift, real-time collaboration and “speedy iteration,” shall be powered by Cerebras’ Wafer Scale Engine 3. The WSE-3 is Cerebras’ third-generation waferscale megachip, decked out with 4 trillion transistors. OpenAI describes the brand new light-weight software as a “each day productiveness driver, serving to customers with speedy prototyping” relatively than the longer, heavier duties that the unique 5.3 is designed for. Spark is at present having fun with a analysis preview for ChatGPT Professional customers within the Codex app.
In a tweet upfront of the announcement, CEO Sam Altman appeared to trace on the new mannequin. “We have now a particular factor launching to Codex customers on the Professional plan later at present,” Altman tweeted. “It sparks pleasure for me.”
In its official assertion, OpenAI emphasised Spark as designed for the bottom doable latency on Codex. “Codex-Spark is step one towards a Codex that works in two complementary modes: real-time collaboration once you need speedy iteration, and long-running duties once you want deeper reasoning and execution,” OpenAI shared. The corporate added that Cerebras’ chips excel at aiding “workflows that demand extraordinarily low latency.”
Cerebras has been round for over a decade however, within the AI period, it has loved an more and more distinguished position within the tech trade. Simply final week, the corporate introduced that it had raised $1 billion in recent capital at a valuation of $23 billion. The corporate has previously announced its intentions to pursue an IPO.
“What excites us most about GPT-5.3-Codex-Spark is partnering with OpenAI and the developer group to find what quick inference makes doable — new interplay patterns, new use instances, and a basically totally different mannequin expertise,” Sean Lie, CTO and Co-Founding father of Cerebras, mentioned in a press release. “This preview is only the start.”
Techcrunch occasion
Boston, MA
|
June 23, 2026
