GPT-5.6 in ChatGPT: Sol, Pro, Plans and Model Picker Guide
GPT-5.6 is generally available, but many ChatGPT users open the app and cannot find a button with that exact name.
That is because GPT-5.6 is a family with three tiers, Sol, Terra, and Luna, while standard ChatGPT conversations expose reasoning levels such as Instant, Medium, High, Extra High, and Pro.
This guide explains which model powers each option, which plans include GPT-5.6, why GPT-5.5 Instant remains the default, and when Sol Pro is worth using.
Quick answer
- GPT-5.6 launched on July 9, 2026 across ChatGPT, Codex, and the OpenAI API.
- Medium, High, and Extra High in standard ChatGPT conversations use GPT-5.6 Sol.
- Pro uses GPT-5.6 Sol Pro for difficult and long-running workflows.
- Instant remains powered by GPT-5.5 Instant.
- Terra and Luna are not selectable in standard ChatGPT conversations, but are available in eligible ChatGPT Work, Codex, and API experiences.
- Rollout is gradual, so an eligible account may not receive access immediately.
In short, GPT-5.6 is usually represented by reasoning levels rather than one dedicated model button in ChatGPT.
What are GPT-5.6 Sol, Terra, and Luna?
| Model | Positioning | Best fit | |---|---|---| | GPT-5.6 Sol | Flagship | Complex reasoning, coding, research, science, design | | GPT-5.6 Terra | Balanced | Everyday agent work and cost-performance balance | | GPT-5.6 Luna | Fast and affordable | High-volume and cost-sensitive workloads |
OpenAI describes the number as the generation and the names as durable capability tiers that may advance on their own cadence. Choosing a GPT-5.6 model therefore requires more than checking the generation number.
Which model powers each ChatGPT option?
| ChatGPT option | Model | Recommended use | |---|---|---| | Instant | GPT-5.5 Instant | Translation, summaries, everyday questions | | Medium | GPT-5.6 Sol | Standard reasoning and file analysis | | High | GPT-5.6 Sol | Research, coding, and complex planning | | Extra High | GPT-5.6 Sol | Difficult tasks that benefit from more reasoning | | Pro | GPT-5.6 Sol Pro | The hardest and longest-running work |
ChatGPT can automatically switch from Instant to Medium for a complex request. Eligible users can turn this behavior on or off in the model-picker configuration.
GPT-5.6 availability by ChatGPT plan
| Plan | Medium / High | Extra High | Sol Pro | |---|---:|---:|---:| | Plus | Included | Not included | Not included | | Pro | Included | Included | Included | | Business | Included | Included | Included | | Enterprise | Included | Included | Included | | Free / Go | Sol not included in standard chat | Not included | Not included |
Free and Go users can still access GPT-5.6 Terra through Codex under the launch availability rules. Business and Enterprise administrators may also control access by workspace role.
Why do I not see GPT-5.6 in ChatGPT?
The most common explanations are a gradual rollout, an ineligible plan, the simplified reasoning-level interface, managed-workspace restrictions, or an app/account refresh issue.
Do not rely on asking the model to identify itself. A language model may infer its answer from conversation context or public information. The model picker and official account entitlements are more reliable.
GPT-5.6 Sol vs Sol Pro
GPT-5.6 Sol is the flagship model for coding, professional knowledge work, research, cybersecurity, science, computer use, and design. Sol Pro allocates more capability to the most difficult and long-running tasks.
Sol Pro is most relevant for multi-source research, large architecture decisions, demanding financial or legal analysis, scientific reasoning, and tasks where failure carries a high cost. It is usually unnecessary for short rewriting, translation, or simple summaries.
What improved in GPT-5.6?
OpenAI emphasizes coding, professional work, science, computer use, cybersecurity, long context, and tool use. Published results include 52.7% on Agents' Last Exam, 88.8% on Terminal-Bench 2.1, and 90.4% on BrowseComp for GPT-5.6 Sol.
Benchmarks do not guarantee the same result on every prompt, but they show that this release focuses on execution and tool-based work rather than only conversational style.
For ChatGPT users, the most visible benefits may include better compliance with multi-part instructions, more coherent long reports, stronger file and web workflows, and more persistent reasoning on coding or research tasks.
Images, search, and current information
OpenAI's latest models support text and image input, text output, multilingual use, and vision. Availability of search, files, computer use, and other tools depends on plan, product surface, and workspace configuration.
The API documentation lists a February 16, 2026 knowledge cutoff for the GPT-5.6 family. Current events still require live search and source verification.
ChatGPT subscription or API?
A ChatGPT subscription is convenient when a person works directly in the app and wants integrated files, search, and product tools. The API is more appropriate for software products, automation, agents, explicit model selection, cost tracking, and routing.
Teams comparing GPT with Claude, Grok, GLM, DeepSeek, and other models can review live model pricing and availability, buy API tokens, and the unified API tutorials. Model support, regional availability, and pricing can change, so check the current platform page.
Practical model-selection advice
- Use Instant for low-risk, short, speed-sensitive tasks.
- Use Medium for most requests that require real reasoning.
- Use High for coding, research, planning, and long documents.
- Use Extra High when the task is difficult and more latency is acceptable.
- Use Pro for important tasks where failure is costly and verification matters.
The highest setting is not automatically the best value. If Medium consistently solves a workload, using Pro for every request may only increase waiting time and consume allowance faster.
Automatic switching and usage limits
Automatic switching is easy to misunderstand. When enabled, ChatGPT can begin with the fast Instant experience and move a sufficiently complex request to Medium. OpenAI states that this automatic switch does not consume the separate allowance used by manually selected reasoning.
Manual selection is more predictable when you already know a task needs sustained reasoning. Examples include reviewing several uploaded files, building a research plan, debugging a complicated failure, or producing a decision memo with explicit evidence requirements.
If a GPT-5.6 reasoning allowance is reached, ChatGPT may make another model available as a fallback until the allowance resets. This does not necessarily mean the account has been restricted. The reset time, when available, is shown in the product.
The practical distinction is:
- let automatic switching handle mixed everyday use
- select Medium or High before a clearly complex task
- reserve Extra High and Pro for work where deeper reasoning justifies the allowance
- verify the selected model again before continuing a long, important conversation after a limit event
A simple evaluation checklist
Rather than judging GPT-5.6 from one impressive answer, compare it on a repeatable set of tasks. Use the same source files and instructions, then record whether the answer follows every requirement, cites the right evidence, avoids unsupported claims, preserves important details, and needs human repair.
For research and knowledge work, measure factual corrections and missing sources. For coding, measure tests passed and regressions introduced. For writing, measure structural edits and time saved. A stronger model is valuable when it reduces accepted-task cost, not merely when it produces a longer response.
Final verdict
GPT-5.6 changes ChatGPT by separating fast everyday responses from deeper professional reasoning. GPT-5.5 Instant remains the default, GPT-5.6 Sol powers the reasoning levels, and Sol Pro addresses the most difficult workflows.
That design uses compute more efficiently, but it also requires users to understand what sits behind each picker option.
For research, coding, file analysis, and complex knowledge work, GPT-5.6 Sol is a meaningful upgrade. For routine conversation, GPT-5.5 Instant may remain the faster and more appropriate choice.
FAQ
Which ChatGPT option is GPT-5.6?
Medium, High, and Extra High use GPT-5.6 Sol. Pro uses GPT-5.6 Sol Pro. Instant uses GPT-5.5 Instant.
Can ChatGPT Plus use GPT-5.6?
Yes. At launch, Plus includes Medium and High access to GPT-5.6 Sol, but not Extra High or Sol Pro.
Why are Terra and Luna missing from standard ChatGPT?
OpenAI primarily exposes Terra and Luna through ChatGPT Work, Codex, and the API. Standard ChatGPT conversations focus on the Sol tier.
Does GPT-5.6 replace GPT-5.5 Instant?
No. GPT-5.5 Instant remains the default for fast everyday responses, with automatic or manual switching to GPT-5.6 Sol for complex work.
Does GPT-5.6 have usage limits?
Yes. Allowances depend on the plan, reasoning mode, and workspace settings. ChatGPT may temporarily fall back to another model after a reasoning allowance is reached.