OpenAI Merges ChatGPT and Codex Token Limits Starting in December
Starting in December, your ChatGPT and Codex usage will drain from the exact same token pool.
Tools · Source: TokenPost
What happened
OpenAI is changing how it meters your AI usage. Starting in December, subscribers on Plus, Pro, and Business plans will share a single token allowance across both ChatGPT and Codex. You no longer get separate buckets for chatting and coding. The two tools will drain the exact same tank.
This change is part of a broader platform shift. OpenAI is folding ChatGPT Work agent features directly into the main interface. Usage is now strictly measured in tokens. Advanced models and complex agent tasks will eat through your budget much faster. Instant chat models are excluded and will not draw from your shared limit.
Users can track their combined remaining balance and reset times in the settings menu. Once you hit the wall, your options are simple. You can wait for a reset, upgrade your plan, or buy extra credits for eligible features. OpenAI has not yet announced the exact launch date in December or the specific token allowances per plan.
Key facts
- December 2026 — When the unified token allowance takes effect
- 5-hour — The limit time frame that applies to Plus and Business plans
- April 2, 2026 — When OpenAI shifted to token-based usage pricing
Why it matters
This alters how developers and founders budget their daily AI compute. A heavy afternoon of generating code with Codex will now eat directly into the allowance you use for writing emails or analyzing documents in ChatGPT. You have to manage one unified pool of compute. Teams that lean heavily on AI for engineering might find their general chat capabilities squeezed.
This signals the end of flat-rate pricing illusions. OpenAI is forcing users to face the actual computational cost of their requests. They recently doubled limits for higher-priced Pro plans, likely to soften the blow of this merger. As AI models handle more complex tasks independently, providers are ensuring heavy users pay for the exact compute they consume.
For builders
Watch your daily token burn rate
Teams relying heavily on Codex will drain their shared pool fast. You need to monitor which activities consume the most tokens to avoid hitting limits mid-day.
Leverage instant chat for basic tasks
Instant chat models do not count toward your combined token budget. Route simple queries to these models to save your premium tokens for complex coding.
Prepare for shifting plan limits
OpenAI has a history of adjusting allowances without much warning. Do not build rigid workflows around current limits because the numbers will likely change again.
My take
OpenAI is finally making you pay for the exact compute you use. Merging chat and code into one meter makes sense for their bottom line, but it means you can no longer game the system with separate buckets. If you code heavily, prepare to hit your limits faster or pay up for more credits.
Original reporting: TokenPost. This is my rewrite and opinion.