OpenAI has released the GPT-5.3-Codex-Spark Research Preview.

February 13th, Today OpenAI released the GPT-5.3-Codex-Spark research preview, a slimmed-down version of GPT-5.3-Codex and OpenAI's first model specifically designed for real-time coding. It supports targeted editing, logic reshaping, or interface optimization, with immediate feedback. Codex-Spark is optimized to run on ultra-low-latency hardware, achieving nearly instant responses—delivering over 1000 tokens per second while still handling real coding tasks effectively.