OpenAI Debuts GPT-5.3-Codex-Spark, a Near-Instant AI for Real-Time Coding

OpenAI has released a research preview of GPT-5.3-Codex-Spark, a lightweight version of its latest Codex system built specifically for real-time coding. The model is designed to deliver rapid responses for tasks such as editing functions, refining logic, or adjusting interfaces inside development tools. According to OpenAI, the model is optimized to feel “near-instant” and can produce more than 1,000 tokens per second when running on ultra-low-latency hardware.At launch, Codex-S