Cerebras has launched its first product powered by its advanced hardware, the GPT-5.3-Codex-Spark, which is specifically designed for real-time coding and operates at over 1,000 tokens per second. This development comes after a significant strategic partnership, where OpenAI agreed in January to purchase up to 750 megawatts of Cerebras compute capacity through 2028, a deal valued at more than $10 billion. Cerebras has filed for an IPO, seeking to capitalize on growing AI demand and the momentum from its partnerships, including this notable collaboration with OpenAI.

OpenAI: OpenAI is an artificial intelligence research organization developing large language models and tools like ChatGPT and Codex for enterprise adoption and developer productivity. It is expanding real-time AI capabilities amid accelerating industry use. OpenAI’s launch of GPT-5.3-Codex-Spark on Cerebras hardware fulfills part of a January compute capacity agreement through 2028.
Cerebras: Cerebras Systems is a Silicon Valley AI infrastructure firm specializing in wafer-scale processors like the Wafer-Scale Engine that integrate massive on-chip memory for AI training and inference, offering an alternative to traditional GPU clusters. The company provides turnkey supercomputers and cloud services to hyperscalers and enterprises. In this news, Cerebras’ hardware powers OpenAI’s newly launched GPT-5.3-Codex-Spark model, enhancing its IPO story built on a prior OpenAI compute deal.
GPT-5.3-Codex-Spark: GPT-5.3-Codex-Spark is OpenAI’s research preview model, a compact version of Codex tuned for real-time coding with features like precise code edits and contextual queries in interactive workflows. It runs exclusively on Cerebras’ low-latency wafer-scale inference hardware. This product represents OpenAI’s first deployment on Cerebras systems, highlighting speed in developer tools.

IPO Momentum: Cerebras filed its S-1 registration in April 2026 for a Nasdaq IPO under ticker CBRS, driven by AI demand and key partnerships.
Inference Edge: Cerebras wafer-scale chips enable ultra-low latency for AI models like GPT-5.3-Codex-Spark, optimized for real-time applications.
Strategic Deal: OpenAI entered a multi-year agreement in January to access Cerebras compute capacity, providing revenue backlog for the chipmaker.