What happened to the GPT-5.3-Codex-Spark usage limit on ChatGPT Pro?
GPT-5.3-Codex-Spark was deprecated on September 14, 2026, and its separate Pro rate limit ended with the research preview. Codex usage now comes out of one plan allowance, sized by Plus, Pro 5x or Pro 20x.
Checked against the lab's own pages on September 20, 2026. Limits change often, so the date matters.
It didn't move, it ended. OpenAI's Codex changelog dates the deprecation to September 14, 2026, and the research preview that carried the separate quota went with it. Anything Codex costs you now comes out of your plan's one allowance.
- Deprecated
- September 14, 2026
- Launched
- February 12, 2026, a research preview for ChatGPT Pro
- Its quota
- its own rate limits, and usage will not count towards standard rate limits
- Pro 5x now
- 25-225 GPT-6 Astra messages per five-hour period
- Pro 20x now
- 100-900 GPT-6 Astra messages per five-hour period
Why Spark had a limit of its own
Spark ran on different hardware. OpenAI put it on Cerebras' Wafer Scale Engine 3 to get "more than 1000 tokens per second", gave it a 128k context window and kept it text-only, and opened it to Pro subscribers while capacity was still being built out.
Because the serving tier was separate, so was the meter.
During the research preview, Codex-Spark will have its own rate limits and usage will not count towards standard rate limits.
The same post warned that the number behind that meter was never fixed: usage was "governed by a separate rate limit that may adjust based on demand during the research preview", with queuing when demand ran high. That's part of why nobody can tell you what the quota was worth.
What the deprecation notice says
GPT-5.3-Codex-Spark was deprecated on September 14, 2026. The research preview is no longer available in the ChatGPT desktop app, Codex CLI, or IDE extension.
OpenAI's instruction is to switch to one of the recommended models and update saved configurations, custom agents and scripts that still name gpt-5.3-codex-spark. For the speed you were using Spark for, the docs point at Fast mode rather than at a replacement model.
What a Pro plan gets instead
Pro starts at $100 a month and comes in two tiers, described on the Codex pricing page as 5x or 20x higher rate limits than Plus. The estimates below are local messages per five-hour period, and OpenAI calls them estimates rather than fixed message limits.
- GPT-6 Astra
- 5-45 on Plus, 25-225 on Pro 5x, 100-900 on Pro 20x
- GPT-5.6 Sol
- 10-100 on Plus, 50-500 on Pro 5x, 200-2,000 on Pro 20x
- GPT-5.6 Luna
- 250-2,000 on Plus, 1,250-10,000 on Pro 5x, 5,000-40,000 on Pro 20x
Local messages and cloud chats draw on the same allowance, weekly limits may also apply, and image generation burns through it 3-5x faster than a turn without one. Fast mode costs more credits too.
Hit the limit mid-task and Codex finishes the turn you're on, subject to fair use. After that Plus and Pro users can buy credits rather than upgrade. Our page on GPT-6 Astra message limits goes through the same table from the Work side.
One more date to keep: GPT-5.5 retires from ChatGPT, ChatGPT Work and Codex on all plans on October 14, 2026, and OpenAI's replacement for it is gpt-5.6-sol. The OpenAI API isn't affected.