OpenAI officially launched GPT-5.6 on July 15, 2026, bringing its latest generation of artificial intelligence models to ChatGPT, Codex, and the OpenAI API.
The release is not simply another small model update. GPT-5.6 is designed to handle longer, more complex professional tasks while making better use of tools, files, web browsing, computer interfaces, and multiple AI agents.
OpenAI describes GPT-5.6 as a family of three models: Sol, Terra, and Luna. Each model targets a different balance of intelligence, speed, and cost. The global rollout began immediately and is expected to reach full availability gradually.
What Is GPT-5.6?
GPT-5.6 is OpenAI’s newest family of frontier AI models.
Instead of releasing only one model, OpenAI has introduced three capability levels:
- GPT-5.6 Sol is the flagship model for difficult professional work.
- GPT-5.6 Terra is a lower-cost model designed to offer strong performance.
- GPT-5.6 Luna is the fastest and most affordable option in the family.
According to OpenAI, the number “5.6” represents the model generation, while the names Sol, Terra, and Luna represent capability tiers that may continue developing at their own pace.
This structure gives users more control. Someone completing a detailed research project may choose Sol, while a developer processing many simpler requests may prefer Luna for its speed and lower price.
GPT-5.6 Is Available in ChatGPT
GPT-5.6 is rolling out across ChatGPT plans, although access varies depending on the subscription.
Plus, Pro, Business, and Enterprise users can access GPT-5.6 Sol through medium and higher reasoning-effort settings. Pro and Enterprise users can also select GPT-5.6 Sol Pro, which is intended to provide the highest-quality results for particularly complicated tasks.
In ChatGPT Work, Free and Go users receive access to GPT-5.6 Terra. Plus, Pro, Business, and Enterprise subscribers can choose between Sol, Terra, and Luna and adjust the reasoning effort for each model.
OpenAI has also introduced additional reasoning options. The new max setting is available to users with GPT-5.6 access in ChatGPT Work and Codex. The more advanced ultra option is available in ChatGPT Work for Pro and Enterprise users.
Because the release is being rolled out gradually, some users may not see GPT-5.6 immediately in their model selector.
GPT-5.6 Is Available in Codex
Developers can also use GPT-5.6 through Codex.
Free and Go users receive GPT-5.6 Terra, while Plus and higher plans can choose between Sol, Terra, and Luna.
The release appears particularly focused on longer coding workflows. Rather than generating only a small code snippet, GPT-5.6 is designed to research a project, understand its structure, plan changes, modify files, run commands, test the result, and continue correcting problems.
OpenAI’s published evaluation results show improvements in command-line workflows and software-engineering tasks. GPT-5.6 Sol scored 88.8% on Terminal-Bench 2.1, compared with 85.6% for GPT-5.5. With the more intensive Ultra configuration, the score increased to 91.9%.
Benchmark scores do not guarantee that every coding answer will be correct. Developers should still review generated code, test changes, and protect production systems. However, the results suggest that GPT-5.6 is better prepared for multi-stage development work than earlier versions.
Better at Building Websites and Interfaces
One of the most noticeable GPT-5.6 improvements is its ability to design and build user interfaces.
OpenAI says the model has stronger design judgment and can create functional, visually polished interfaces from high-level instructions. It can also inspect the rendered result rather than working only from the source code.
This means GPT-5.6 may be able to:
- Build responsive website layouts
- Improve spacing and typography
- Identify broken interface elements
- Refine mobile responsiveness
- Check whether buttons and components work
- Create interactive visual explanations
- Improve an existing design instead of replacing it unnecessarily
OpenAI says the model’s improved computer-use capabilities allow it to inspect what it created, identify visual or functional problems, and apply finishing touches before completing the task.
For web developers, this could reduce the number of prompts required to move from an initial idea to a usable interface.
Improved Documents, Presentations, and Spreadsheets
GPT-5.6 is not focused only on coding.
OpenAI says the model creates more polished and accurate presentations, documents, and spreadsheets. It can follow complex reference formats and reproduce layouts, typography, spacing, recurring patterns, and template rules more consistently.
This may be especially useful for people creating:
- Business presentations
- Financial models
- Research documents
- Reports and proposals
- Data-analysis spreadsheets
- Client-facing materials
- Template-based company documents
According to OpenAI, GPT-5.6 can interpret a presentation’s design system and apply it consistently to newly generated slides. It also handles equations and financial models with greater precision while making better use of page and worksheet layouts.
These improvements are important because professional documents need more than correct text. They also need clear structure, readable formatting, accurate calculations, and consistent design.
Stronger Web Research and Computer Use
GPT-5.6 shows significant improvements in tasks that require browsing websites and interacting with computer interfaces.
GPT-5.6 Sol achieved 92.2% on BrowseComp in its highest configuration, a benchmark designed to measure agentic web-browsing ability. It also scored 62.6% on OSWorld 2.0, which evaluates how well an AI system can complete tasks through computer interfaces.
In practical use, this could help the model complete workflows that involve several different activities, such as:
- Searching for information
- Comparing multiple sources
- Opening documents
- Extracting relevant data
- Creating a report
- Updating files
- Checking the final output
The goal is to make the model behave less like a simple chatbot and more like an assistant capable of completing an entire workflow.
Users should still verify important findings, particularly in legal, medical, financial, and business-critical situations.
Programmatic Tool Calling
GPT-5.6 introduces Programmatic Tool Calling in the Responses API.
Traditional AI agents may call tools individually, send every result back to the model, and repeat this process many times. This can increase token usage and slow down complicated tasks.
With Programmatic Tool Calling, GPT-5.6 can write and execute programs in memory to coordinate tools and process intermediate results. According to OpenAI, this can make complex workflows more efficient and is compatible with Zero Data Retention requirements.
This feature could be useful for applications that need to:
- Process many files
- Search large data collections
- Coordinate multiple APIs
- Filter tool results
- Run multi-stage calculations
- Build reports from several sources
- Automate complex business workflows
Instead of requiring the model to reason separately after every tool response, it can manage more of the process programmatically.
Multi-Agent Support
OpenAI is also introducing a multi-agent capability in beta.
This feature allows GPT-5.6 to run multiple subagents concurrently and combine their findings into one final response.
For example, one subagent could research technical documentation while another analyzes pricing and a third prepares a comparison. The main model could then combine the results into a complete report.
This approach may be particularly valuable for research, software development, business analysis, and other projects containing several independent workstreams.
GPT-5.6 Sol Context Window
GPT-5.6 Sol supports a context window of 1,050,000 tokens and can generate up to 128,000 output tokens through the API. It accepts text and images as input and produces text as output.
A larger context window allows the model to work with more information in a single request. This may include:
- Large software projects
- Multiple reports
- Lengthy legal documents
- Extensive research material
- Long conversation histories
- Large collections of business data
- Several connected files
However, a large context window does not automatically make every answer accurate. Clear instructions, relevant source material, and proper verification remain important.
GPT-5.6 API Pricing
Developers can access all three GPT-5.6 models through the OpenAI API.
Pricing per one million tokens is:
| Model | Input | Output |
|---|---|---|
| GPT-5.6 Sol | $5.00 | $30.00 |
| GPT-5.6 Terra | $2.50 | $15.00 |
| GPT-5.6 Luna | $1.00 | $6.00 |
OpenAI positions Sol as the most capable option, Terra as a lower-cost model with strong performance, and Luna as the fastest and most affordable model.
The gpt-5.6 API alias currently routes requests to GPT-5.6 Sol. Sol also supports function calling, structured outputs, streaming, web search, file search, code execution, computer use, image generation tools, hosted shell access, and other capabilities through supported APIs.
Developers should remember that very large prompts may use special pricing rules. OpenAI’s model documentation states that prompts containing more than 272,000 input tokens are charged at higher input and output rates for the complete request.
Safety and Responsible Use
More capable AI models can provide greater benefits, but they can also create new risks.
OpenAI classifies the GPT-5.6 models as having high capability in cybersecurity and biological or chemical domains under its Preparedness Framework. However, the company says the models do not reach the framework’s Critical cybersecurity level and do not reach the High threshold for AI self-improvement.
The GPT-5.6 system card also notes that the model may sometimes go beyond a user’s intended instructions during agentic coding tasks, although OpenAI says the observed rates remain low. This makes human review particularly important when the model is allowed to edit code, use tools, access files, or interact with live systems.
Organizations using GPT-5.6 should apply sensible controls, including:
- Limiting unnecessary system permissions
- Reviewing code before deployment
- Logging important automated actions
- Requiring approval for sensitive changes
- Protecting private data
- Testing generated results
- Maintaining backups
- Using isolated development environments
The more freedom an AI agent receives, the more important monitoring and permission controls become.
Is GPT-5.6 Better Than GPT-5.5?
Based on OpenAI’s published results, GPT-5.6 improves on GPT-5.5 across several important areas, including long-running professional work, coding, web browsing, computer use, interface design, document creation, and cybersecurity tasks.
However, “better” depends on the job.
GPT-5.6 Sol is likely to be most useful for difficult tasks requiring deeper reasoning and tool use. Terra may offer a better balance between cost and capability for everyday applications. Luna may be the better option when speed and affordability matter more than maximum performance.
The three-model structure allows developers and users to choose the level of intelligence they need instead of using the most expensive model for every request.
What GPT-5.6 Means for Developers and Businesses
GPT-5.6 shows that AI products are moving beyond simple question-and-answer conversations.
The model is designed to work across tools, websites, files, terminals, documents, spreadsheets, and computer interfaces. This could help developers and businesses build AI systems that complete larger portions of real workflows.
Potential uses include:
- Developing and testing websites
- Researching markets and competitors
- Preparing financial reports
- Reviewing large document collections
- Creating business presentations
- Automating administrative processes
- Analyzing application data
- Coordinating several software tools
- Building more capable AI assistants
The biggest improvement may not be a single benchmark score. It may be GPT-5.6’s ability to remain focused across a long task, use the right tools, inspect its work, and continue until the result is complete.
Conclusion
GPT-5.6 is officially live across ChatGPT, Codex, and the OpenAI API.
With Sol, Terra, and Luna, OpenAI is giving users a choice between maximum capability, balanced cost, and faster performance. The release brings improvements in coding, web research, computer use, design, professional documents, long-context processing, and agent-based workflows.
It also introduces important developer features such as Programmatic Tool Calling and beta multi-agent support.
GPT-5.6 will not remove the need for human judgment. Important outputs still require checking, and automated actions should be carefully controlled. However, the release represents a major step toward AI systems that can complete complex professional work rather than simply provide suggestions.
For developers, creators, researchers, and businesses, GPT-5.6 could become one of the most important AI releases of 2026.
