As it battles rival Anthropic for the enterprise market, OpenAI introduced a new coding model powered by an advanced AI chip from startup Cerebras.
OpenAI released a new coding model, GPT-5.3-codecs-spark, in research preview on February 12. The generative AI vendor designed the model for real-time coding as a smaller version of the GPT-5.3 codec, which OpenAI released earlier this month for computer use and coding tasks.
Codex-Spark is the first OpenAI model not to use Nvidia’s hardware, running entirely on Cerebras Wafer-Scale Engine 3 chips.
The release of Codex-Sparks comes as both Cerebras and OpenAI Enterprises are trying to prove their worth against their competitors.
For Cerebrus, a specialized semiconductor and cloud computing startup, Codex-Spark’s effectiveness could show potential customers that its larger AI chips and wafer-scale designs can be just as valuable as Nvidia’s market-dominating GPUs. Meanwhile, for OpenAI, improving its coding model could boost its credibility in a market dominated by Anthropic, especially considering that the creator of the popular cloud model recently pitched another model. $30 billion And it’s putting $20 million into a new super PAC to counter OpenAI’s super PAC.
Benefits and Opportunities
GPT-5.3-Codex-Spark heavily focused on real time Coding. OpenAI said the model works with codecs for tasks like targeted editing, reshaping logic, and getting the job done in a moment.
As a smaller model, Codex-Spark is easier to support and more cost-effective for developers looking for immediate support; However, it is limited in some features, said Lian Jae Su, an analyst at Omdia, a division of Informa TechTarget. For example, context window Has only 128k and supports text-only signals.
“For some use cases, this is probably sufficient, and it’s really cleverly designed to target a specific segment of the coding population,” Su said. Beginner coders or those looking for real-time coding assistance may find Codex-Spark most attractive, he said.
In addition, the use of OpenAI Cerebra’s Wafer-Scale Engine Infrastructure This could represent an opportunity for other AI hardware vendors like Grok or Tenorient, which specialize in application-specific integrated circuits, or ASICs.
“Making these services available and running on AI ASICs, especially those designed specifically for inference, creates this business model for similar players in the market,” Su said. He added that, with Cerebra’s high-throughput inference chips, Codex-Sparks supports low-latency, real-time inference.
Challenges and ventures
However, it’s clear that much of the engineering still needs to be configured on the backend to utilize OpenAI’s Cerebras GPUs instead of Nvidia’s. Su said the shift from GPUs to wafer-scale engines will require significant reconfiguration, porting, and codebase conversion outside of Nvidia’s architecture.
However, if OpenAI is successful with Cerebra, other AI model makers may decide to try other hardware options in the future, Su said.
For enterprises, the hardware behind the system matters less than whether the product functions effectively.
“They only care about whether it works for them or not, the accuracy, the responsiveness, whether it’s really like they said it was, really low latency,” he said.
