Introducing GPT-5.3-Codex-Spark
OpenAI
OpenAI just dropped GPT-5.3-Codex-Spark, a coding model built for real-time speed. It generates code 15x faster than before—Pro users get early access now.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI has a new coding model out, and the headline number is speed. GPT-5.3-Codex-Spark generates code 15 times faster than its predecessors, which OpenAI is billing as its first genuinely real-time coding model. That's not a marginal bump — that's the difference between waiting for a response and watching code appear as fast as you can read it.
The model ships with a 128k context window, enough to hold a sizable codebase or a long, messy conversation history without losing the thread. For developers who've grown used to models forgetting what happened three files ago, that context length matters almost as much as the raw speed.
Right now this is a research preview, and only ChatGPT Pro subscribers get to touch it. That's a familiar OpenAI move: put the fastest, newest thing behind the priciest tier first, gather feedback, then figure out wider rollout later. No word yet on API access or pricing for the model itself, and OpenAy hasn't detailed exactly what architectural changes produced the speed gain.
Still, the framing here is telling. OpenAI isn't pitching Codex-Spark as smarter than its other coding models — it's pitching it as faster, full stop. And in a market where Anthropic, Google, and a growing pile of open coding models are all racing on capability, speed is starting to look like the next battlefield.
My take — AI-written commentary, not fact-checked reporting
Fifteen times faster is the kind of number that sounds great in a blog post and means nothing until real devs bang on it — I want to see it write a Rust borrow-checker fix before I believe the hype. Also, gating this behind ChatGPT Pro instead of the API tells you who OpenAI actually wants feedback from right now: power users with money, not the open-source crowd that usually finds the sharp edges first.”}}
Read more about this at: OpenAI