What changed
OpenAI released a 128k, text-only research preview of GPT-5.3-Codex-Spark for low-latency coding in Codex. Served on Cerebras Wafer Scale Engine 3, it was reported at more than 1,000 tokens per second and was available to ChatGPT Pro users with limited API design-partner access.
Why it matters
Interactive editing and long-running autonomous work are separating into different serving paths, models and hardware rather than forcing one model to optimize for both.
What to do
Use the fast tier for narrow edits, logic reshaping and rapid interface iteration; keep deeper refactors and autonomous validation on a longer-horizon model. Explicitly ask Spark to run tests when the change needs them.
Caveat
This was a limited research preview with separate rate limits. Speed and benchmark claims are first-party, the model was text-only, and its lightweight default did not automatically run tests.