Cognition has added a way to run GPT models inside Devin without spending your Devin quota on them. In its docs, the company calls this token sharing: you link a personal ChatGPT subscription, turn on "Use your ChatGPT plan," and GPT model requests Devin makes on your behalf are billed to ChatGPT instead. Cognition announced the change on X at 06:52 Beijing time on October 10, saying any personal ChatGPT plan could connect and that all GPT usage would draw from that plan's quota. The docs are a notch more precise.
On the ChatGPT side, Go, Plus and Pro work; free ChatGPT accounts don't. On the Devin side, token sharing is limited to the self-serve Pro, Max and Teams plans — Free and Enterprise are out. On Teams, each person links their own ChatGPT plan, and only the sessions they own can draw from it.
According to the docs, the ChatGPT plan covers every GPT model available in Devin except fast and priority variants. Fusion and Adaptive setups still mix in other models, and those keep drawing from your Devin quota as usual. The GPT slice is billed only to ChatGPT; it doesn't also eat Devin quota or on-demand credits.
When the ChatGPT plan hits its limit, work doesn't stop. Devin falls back to normal routing, bills the GPT portion to your Devin quota, and then to on-demand credits once that quota is gone. After ChatGPT resets, GPT usage goes back to the ChatGPT plan. You can unlink the account or flip the switch off anytime in Settings.
Tibo (@thsottiaux) quoted Cognition's post at 12:33 Beijing time the same day with the line "Your ChatGPT subscription is now also a Devin subscription." That is his paraphrase, not Cognition's wording. The docs are clearer: only the GPT part moves over, and only on the plans listed above.
ChatGPT plan allowances themselves are still opaque in places. Our guide "OpenAI Keeps Codex's Weekly Caps Vague: What Each ChatGPT Plan Actually Gets" covers what OpenAI does and doesn't publish about those caps.
