OpenAI extends Cursor’s GPT access past November 12. How to keep using GPT there

OpenAI extends access past November 12

On August 28, 2026, two weeks after SpaceX completed its $60 billion purchase of Cursor, OpenAI said it had notified SpaceX “that we intend to wind down our contract providing OpenAI models to Cursor, with a proposed shutoff date of November 12, 2026.” Its reason: “we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk’s companies violating contracts.”

On November 12, the day itself, OpenAI added a line to the top of that post: “Update on November 12, 2026: We’ve temporarily extended Cursor’s access to our models beyond November 12. We will provide updates as available.” That’s all it says. There’s no new end date, no reason, and no word on how long “temporarily” means. As of November 13:

• OpenAI’s help page hasn’t caught up. Its page on Cursor, last updated in late October, still says it has “proposed a transition period of November 12, 2026,” that “Cursor may choose to end access sooner,” and that it will share the official termination date “as soon as it is confirmed between both companies.”
• Cursor still lists GPT. Its models and pricing page lists GPT-5.6 Sol, Terra and Luna among its third-party models, with no note about November 12 or the extension, and its help page still recommends “Claude Opus and GPT-5.6 Sol” for complex, multi-step tasks. The only dated note beside an OpenAI model is GPT-5.6 Sol’s “Promotional pricing through November 21, 2026,” which is about its price; neither company has tied that date to the contract.
• Cursor hasn’t said anything in public that we could find. Its last word is from its CEO, Michael Truell, the night of the announcement: “OpenAI models serve about 5% of Cursor user traffic, and we’re speaking with the OpenAI team to resolve this,” CNBC reported.

So GPT works in Cursor for now, and the cutoff could still come on short notice. Treat the extension as time to set up a fallback, not as a reprieve.

One part is settled. OpenAI said it would hold the cancellation “to the latest date we can while not providing future models to Cursor,” and it has. OpenAI’s current family, GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna, isn’t in Cursor and won’t be under these terms. The Terra in Cursor is the older GPT-5.6 Terra.

How to tell whether GPT still works for you

1. Look at the model picker. In Chat or Agent, check for GPT-5.6 Sol, Terra and Luna. One catch: if you’ve added an OpenAI key under Cursor Settings > Models, Cursor’s API keys help says it “now uses your key for that provider’s models,” so GPT can stay in your picker after Cursor’s own supply ends, billed by OpenAI.
2. Check Auto. On Teams and Enterprise, Cursor Router picks the model for each request in its Balance and Intelligence modes. Cursor’s help page still lists GPT-5.5 among the four models it routes across, with Claude Opus 5, Grok 4.6 and Claude Fable 5, and says: “Recommended: GPT-5.5 and Claude Opus 5. Blocking one reduces routing quality. Blocking both disables the router.” It hasn’t said what replaces GPT-5.5.
3. Check your subagents. A custom subagent can name its model in the model field of its YAML frontmatter; gpt-5.6-sol is one of Cursor’s own examples. Cursor’s subagents docs say it “falls back to a compatible model” when an admin has blocked that model or it “isn’t available on your current plan.” They don’t say what happens when a model leaves Cursor altogether.
4. Check Cloud Agents and Automations. These run on models Cursor supplies, and OpenAI’s page says your own key or gateway can’t power them. Any set to GPT will need another model once the cutoff lands.

Who it affects

How you reach GPT in Cursor depends on your plan. New customers have been on The X Subscription since October 12, and Cursor’s models and pricing page, API keys help and legacy plans page set out each one:

Plan	GPT from Cursor, until the cutoff	Your own OpenAI API key
Plus, $8 a month	Not available	Not available
Premium, Super or Ultra, $30 to $200	At API list price, from prepaid credit	Yes, billed by OpenAI
Teams, $30 to $200 a seat	At API list price, plus Cursor’s token rate	Yes, plus Cursor’s token rate
Legacy Pro, Pro+ or Ultra	From the Other Models pool	Yes, billed by OpenAI

Free, like Plus, includes only Cursor’s own models: Grok 4.7, 4.6 and 4.5, and Composer 2.5. Cursor’s token rate on Teams is $0.25 per million tokens; models whose cache-read price is under $0.10 per million, such as GPT-5.6 Luna, are exempt.

When the cutoff lands, the middle column ends for everyone. Nothing announced touches Claude, Gemini, Muse Wonder, Grok or Composer. Anthropic’s Tom Brown wrote the night OpenAI announced it that “We’ll continue to increase compute to support Claude models in Cursor.”

Four ways to keep using GPT

1. Your own OpenAI API key, in Cursor. Paste it under Cursor Settings > Models and enable the model. OpenAI bills you at API prices: $2 in and $10 out per million tokens for GPT-6.1 Sol, $10 and $50 for GPT-6 Astra, $1 and $5 for GPT-6.1 Terra, and $0.10 and $0.50 for GPT-6 Luna. OpenAI’s page says a key works only in Cursor’s local Chat and Agent, not in “Cursor Tab, Auto, Cloud or Background Agents, Automations, the Cursor CLI, or Cursor’s API and SDK.” Cursor’s own help describes its OpenAI support as “Standard, non-reasoning chat models” and says the model picker shows which are available, so check that the one you want appears. Cursor’s zero data retention doesn’t apply to requests on your own key.
2. OpenAI’s Codex extension, inside Cursor. The Codex IDE extension lists Cursor among the editors it installs in, and signs in with ChatGPT or an API key. It’s a separate agent in its own panel; Cursor’s Agent, Tab and Cloud Agents don’t change. It’s the only option here that runs on a ChatGPT plan; as OpenAI puts it, “A ChatGPT subscription does not include OpenAI API usage.” Codex pricing puts Codex in the IDE extension on Plus, with GPT-6.1 Sol and GPT-6 Luna; for Free and Go it lists only the desktop app.
3. A gateway. OpenAI also names Amazon Bedrock, Azure and other OpenAI-compatible gateways, set up in the same Models settings. They have the same limits as a key: local Chat and Agent only.
4. Another model in Cursor. On Premium and above, from prepaid credit, Claude Opus 5.5 costs $4 in and $20 out per million tokens, Claude Fable 5.2 $8 and $40, and Claude Sonnet 5.5, Gemini 4 Argon and Gemini 4 Carbon $2 and $10. Grok 4.7 and Composer 2.5 come out of your plan’s included usage.

What to do while the extension lasts

• Set up your fallback before you need it. If GPT matters to your work in Cursor, add your own OpenAI key or install the Codex extension now, and run one real task on it. OpenAI hasn’t promised notice before access ends.
• Find everything that names a GPT model. The picker is the obvious place. Less obvious are subagents, Cloud Agents and Automations, which a key can’t take over.
• Recheck the X Subscription without GPT. Legacy Pro, Pro+ and Ultra include third-party models through the Other Models pool; The X Subscription doesn’t, and Cursor says the switch “is permanent.” Once GPT leaves Cursor, that pool can’t buy it, but it still covers Claude and Gemini.
• If you pay for ChatGPT, use it. Codex, in the extension, the CLI or OpenAI’s app, runs GPT-6 models on the plan you already have. From December, Plus, Pro and Business share one usage limit across ChatGPT and Codex.

Using GPT outside Cursor

Enjoy runs Codex on your ChatGPT plan beside Claude Code, Grok Build and the other labs’ agents, each signed in to its own account. Start a job with one, switch the conversation to Codex when you want GPT-6.1 Sol, and Codex gets the messages it missed. Enjoy is a monthly subscription, with the first seven days free (pricing). It doesn’t put OpenAI’s models back into Cursor, and it doesn’t add GPT to a Grok or Claude account.

What neither company has said

• The new end date, how long the temporary extension runs, or why OpenAI granted it.
• Whether talks could still change the outcome. OpenAI’s original post still says it will hold the cancellation “to the latest date we can while not providing future models to Cursor.”
• Which OpenAI models a key reaches in Cursor’s picker, and whether GPT-6 is among them.
• What Cursor Router uses in place of GPT-5.5.
• What happens on the day to a conversation, subagent, Cloud Agent or Automation set to GPT.

We’ll update this memo when either company sets a date or the models leave Cursor.