Cursor Adds GPT-5.6 Sol, Terra, and Luna
Cursor added OpenAI's GPT-5.6 Sol, Terra, and Luna on July 9, 2026, and as of October 2026 they are still the newest OpenAI models Cursor lists. Cursor offers no GPT-6 model, and OpenAI has proposed ending Cursor's direct access to its models on November 12, 2026. Sol scores 67.2% on CursorBench 3.2.
Status (October 9, 2026): GPT-5.6 is still the newest OpenAI family in Cursor. Cursor lists no GPT-6 model, and OpenAI has proposed ending Cursor's direct access to its models on November 12, 2026.
Cursor added the GPT-5.6 model family -- Sol, Terra, and Luna -- on July 9, 2026, selectable directly from the model picker (Cursor announcement). Three months on, they remain the newest OpenAI models on Cursor's models and pricing page. OpenAI's newer GPT-6.1 Sol and GPT-6 Sol and Luna are not offered in Cursor, and OpenAI has said it will not provide future models to Cursor under its current contract (our coverage of the proposed cutoff).
Key Takeaways
- Three models: Sol (flagship, hardest coding and agent work), Terra (balanced, everyday development), Luna (fastest, cheapest).
- Benchmark: Cursor says GPT-5.6 Sol scores 67.2% at Max effort on CursorBench 3.2.
- Pricing in Cursor: per million tokens, Sol $4 input / $20 output, Terra $2 / $12, Luna $0.20 / $1.20, all drawing from the Other Models pool (Cursor models and pricing).
- No GPT-6 in Cursor: Cursor's model list stops at GPT-5.6 for OpenAI as of October 9, 2026.
- Proposed cutoff: OpenAI has proposed November 12, 2026 to end direct Cursor access to its models; the date is not final (Tom's Guide).
What Cursor Says Is Available
Cursor's announcement includes three GPT-5.6 variants: (Cursor announcement)
- GPT-5.6 Sol: described as the flagship model for difficult coding problems and complex agent workflows. (Cursor announcement)
- GPT-5.6 Terra: described as a balanced model for everyday development work. (Cursor announcement)
- GPT-5.6 Luna: described as the fastest and most cost-efficient option for quick edits and iteration. (Cursor announcement)
Cursor says you can pick GPT-5.6 Sol, Terra, or Luna directly from the model picker. (Cursor announcement)
Pricing in Cursor
Cursor's models and pricing page lists these per-million-token rates as of October 9, 2026. All three draw from the third-party Other Models pool and require Max Mode on legacy request-based plans.
| Model | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| GPT-5.6 Sol | $4.00 | $5.00 | $0.40 | $20.00 |
| GPT-5.6 Terra | $2.00 | $2.50 | $0.20 | $12.00 |
| GPT-5.6 Luna | $0.20 | $0.25 | $0.02 | $1.20 |
Fast mode is available for each at 2x pricing, and input above 272k tokens bills at 2x. Cursor's table notes promotional pricing on Sol through November 21, 2026.
A Benchmark Datapoint (CursorBench)
Cursor says GPT-5.6 Sol scored 67.2% at Max effort on CursorBench 3.2. (Cursor announcement)
How to Use This Update
If you want the best agent planning and persistence for complex work, start by trying Sol on one of your longer-running tasks (multi-file changes, tool-heavy debugging, or repo-wide refactors), then compare it against your current default model on the same prompt. Cursor's Sol model page says it "keeps going through multi-hour agent sessions instead of stopping early," but notes it can over-use subagents on mid-sized tasks.
For routine agent workflows, Cursor's Terra page positions it as faster and cheaper than Sol. If you're optimizing for throughput (lots of smaller edits), Cursor's Luna page recommends it for "high-volume, latency-sensitive, or cost-sensitive agent work."
For prompting advice on this family, see our GPT-5.6 prompting guide.
What Happens After November 12
OpenAI plans to wind down the contract that supplies its models directly to Cursor, with November 12, 2026 proposed and not yet confirmed. Cursor says OpenAI models serve about 5% of its user traffic and that it will share more "in the coming weeks around continuity of access to OpenAI models" (Cursor forum). If you rely on GPT-5.6 for local Chat or Agent, set up your own OpenAI API key or the Codex extension; those routes do not cover Tab, Auto, Cloud Agents, or automations. The full options are in our OpenAI and Cursor cutoff guide. If you want GPT-6.1 Sol itself, you need OpenAI's own tools or API; see the GPT-6.1 Sol launch coverage.
Corrections and Updates
- October 9, 2026: This page presented GPT-5.6 as the OpenAI choice in Cursor without context; OpenAI has since released GPT-6 models that Cursor does not offer, and OpenAI has proposed ending Cursor's direct access to its models on November 12, 2026.
Sources
- Cursor forum, "GPT-5.6 now available in Cursor" (July 9, 2026), read 2026-10-09: https://forum.cursor.com/t/gpt-5-6-now-available-in-cursor/165312
- Cursor docs, "Models and pricing", read 2026-10-09: https://cursor.com/docs/models-and-pricing
- Cursor docs, GPT-5.6 Sol, Terra, and Luna model pages, read 2026-10-09: https://cursor.com/docs/models/gpt-5-6-sol, https://cursor.com/docs/models/gpt-5-6-terra, https://cursor.com/docs/models/gpt-5-6-luna
- Cursor forum, staff replies on OpenAI model access (August 29 and September 7, 2026), read 2026-10-09: https://forum.cursor.com/t/i-heard-openai-is-about-to-cut-off-gpt-access-for-cursor-is-this-true/169876
- Tom's Guide, "OpenAI is leaving Cursor in November", read 2026-10-09: https://www.tomsguide.com/ai/openai-is-leaving-cursor-in-november-here-are-your-3-options
Read next
Related coverage- OpenAI
OpenAI Plans to End Direct Cursor Model Access on November 12
OpenAI plans to end Cursor's direct access to OpenAI models. November 12, 2026 is proposed, and as of October 9 neither company has confirmed it. Only OpenAI models picked inside Cursor are affected, about 5% of its traffic. Local Chat and Agent can keep OpenAI through your own API key or the Codex extension; Tab, Auto and Cloud Agents cannot.
- OpenAI
GPT-6.1 Sol: Astra-Level Coding at a Fifth of the Price, With Cost at Every Effort Level
OpenAI released GPT-6.1 Sol on September 29, 2026, at $2 input and $10 output per million tokens, a fifth of GPT-6 Astra's price, with cached input at $0.10. On OpenAI's charts it edges past Astra's best DeepSWE score at high effort and trails Astra by 2.1 points on OSWorld offline. The API model ID is gpt-6.1-sol.
- OpenAI
OpenAI Previews GPT-5.6: Sol, Terra, and Luna in Limited Preview
OpenAI previewed the GPT-5.6 family in June 2026 with trusted partners: Sol, a frontier flagship it called a step function better than GPT-5.5; Terra, competitive with GPT-5.5 at 2x lower cost; and Luna, its most cost-efficient model. All three became generally available on July 9, 2026, and GPT-6 has since superseded them.
- OpenAI
Prompting GPT-5.6: Message Roles, Effort, and Agentic Prompts That Work
GPT-5.6 rewards precise, explicit prompts: structure developer messages as identity, instructions, examples, then context, keep stable content first for prompt caching, and pick reasoning effort deliberately (xhigh for complex multi-step work). Move saved prompt objects into code before OpenAI shuts down v1/prompts on November 30, 2026.
Frequently Asked Questions
Which GPT-5.6 models did Cursor add?
Cursor added the GPT-5.6 family on July 9, 2026: GPT-5.6 Sol for difficult coding and complex agent workflows, GPT-5.6 Terra for everyday development, and GPT-5.6 Luna for fast, low-cost edits. All three are selectable from Cursor's model picker.
Can I use GPT-6 or GPT-6.1 Sol in Cursor?
Not as a built-in model. As of October 9, 2026, Cursor's models page lists GPT-5.6 Sol, Terra, and Luna as its newest OpenAI models and no GPT-6 model. OpenAI has said it will not provide future models to Cursor under its current contract.
Will GPT-5.6 keep working in Cursor after November 12, 2026?
Possibly not through Cursor's own access. OpenAI has proposed November 12, 2026 to end direct Cursor access to its models, and the date is not final. Local Chat and Agent can keep OpenAI models with your own API key; Tab, Auto, Cloud Agents, and automations cannot.
Which GPT-5.6 model should I pick in Cursor?
Cursor's docs position Sol for the hardest long-running agent work, Terra for routine agentic coding at well under Sol's price, and Luna for high-volume, latency-sensitive, or cost-sensitive work such as subagents. Sol scores 67.2% at Max effort on CursorBench 3.2.