GPT‑5.3‑Codex‑Spark is here

• February 12, 2026 at 23:57 IST
• 1 min read

• openai
• codex
• coding

OpenAI just launched a faster version of GPT-5.3-Codex and named it GPT‑5.3‑Codex‑Spark. As they mention in this post by Sam, it outputs more than 1000 tokens per second. But they also mention "smaller" as you see here:

Today, we’re releasing a research preview of GPT‑5.3-Codex-Spark, a smaller version of GPT‑5.3-Codex, and our first model designed for real-time coding.

And as you can see in the chart here, the new Spark model is slightly less capable that GPT-5.3-Codex but crazy faster than that.

I still haven't gotten access to the model, but still reading about it and will keep updating this page when I get the access and/or learn more about it.