OpenAI Launches GPT-5.3-Codex-Spark on Cerebras Hardware — First Production Deployment Away from NVIDIA

Summary

OpenAI released GPT-5.3-Codex-Spark on February 13, a lightweight version of its GPT-5.3-Codex agentic coding model optimized for real-time, interactive coding workflows. The model delivers over 1,000 tokens per second on ultra-low-latency hardware, operates with a 128k context window (text-only), and is designed for rapid iteration — targeted edits, logic adjustments, and interface refinements with near-instantaneous feedback. The most strategically significant detail is the hardware: Codex-Sp