On February 12, 2026, OpenAI unveiled its new coding model, GPT-5.3-Codex-Spark, marking a significant development in artificial intelligence technology. This release is part of an ongoing effort by OpenAI to reduce dependency on Nvidia hardware, a move that is indicative of a larger shift in the AI industry. The Codex-Spark model, designed to run on Cerebras chips, is reported to operate at a speed 15 times faster than its predecessor, capable of delivering over 1,000 tokens per second for coding tasks. Despite the impressive speed, OpenAI has not provided independent validation for these numbers, raising questions about the credibility of the reported performance. The growth and development of AI coding agents have been notable over the past year, as various tools have shown increased utility for developing software prototypes and applications, further demonstrating the importance of efficient coding systems. This model is part of a wider strategy by OpenAI, which has signed major deals with companies like AMD and Amazon, and is developing its own AI hardware for future projects. The collaboration with Cerebras also shows a commitment to exploring alternatives to Nvidia’s dominance in the AI chip market, driven by dissatisfaction with the performance of certain Nvidia chips in executing AI inference tasks. The move away from Nvidia is seen as both a technological advancement and a strategic decision that could reshape the competitive landscape of AI development.