OpenAI launches GPT-5.3-Codex-Spark on Cerebras chips — marks AI giants first production deployment away from Nvidia

OpenAI has launched GPT-5.3-Codex-Spark on Cerebras chips, marking its first AI model deployment away from Nvidia. This new model is optimized for coding tasks, offering high throughput on ultra-low latency hardware. OpenAI plans to deploy 750 megawatts of Cerebras-backed compute by 2028 for low-latency inference, complementing Nvidia's role in training infrastructure. Despite the partnership with Cerebras, OpenAI still views Nvidia as foundational to its training and inference stack, with plans to deploy AMD chips and collaborate with Broadcom for custom AI accelerators.