OpenAI Deploys GPT-5.3-Codex-Spark On Cerebras Accelerators

On Thursday, OpenAI unveiled GPT-5.3-Codex-Spark, its first model designed to run on Cerebras Systems' CS3 waferscale accelerators under a $10 billion contract. The lightweight, text-only model offers a 128,000-token context window and can generate responses at more than 1,000 tokens per second using Cerebras' SRAM-packed chips. Spark is available in preview to Codex Pro users and select API partners.