OpenAI sidesteps Nvidia with unusually fast coding model on plate-sized chips

February 13, 2026

TL;DR

OpenAI released GPT-5.3-Codex-Spark, its first production AI model running on non-Nvidia hardware (Cerebras chips).
The new model delivers code at over 1,000 tokens per second, approximately 15 times faster than its predecessor.
Codex-Spark is a research preview available to ChatGPT Pro subscribers and aims to accelerate coding tasks.
This release is part of OpenAI's broader strategy to diversify its hardware suppliers away from Nvidia.
The model runs on Cerebras' Wafer Scale Engine 3 and represents the first product from OpenAI's partnership with Cerebras.
OpenAI has been actively seeking alternative hardware, including deals with AMD and Amazon, and designing its own chips.

Continue reading
the original article