GPT-5.6 Sol vs Terra vs Luna
OpenAI's GPT-5.6 family introduces three capability tiers for complex professional work, affordable everyday AI and fast, high-volume tasks. Here is a complete guide to their features, pricing, ChatGPT availability, API access and best use cases.
OpenAI officially launched the GPT-5.6 model family on July 9, 2026. Instead of offering only one standard model, the company divided GPT-5.6 into three persistent capability tiers: Sol, Terra and Luna.
Sol is designed for the most complicated professional and technical assignments. Terra targets users who need strong performance at a lower operating cost. Luna is intended for fast, affordable workloads where response speed and scale matter more than maximum reasoning power.
GPT-5.6 is available through ChatGPT, ChatGPT Work, Codex and the OpenAI API, although the exact model choices and reasoning settings depend on the user's subscription and product.
- Sol: OpenAI's flagship GPT-5.6 model.
- Terra: Lower-cost model with performance positioned close to GPT-5.5.
- Luna: The fastest and most affordable GPT-5.6 tier.
- API prices: From $1 input and $6 output per one million tokens.
- New modes: Max reasoning and multi-agent Ultra workflows.
- Major focus: Coding, research, documents, design, cybersecurity and agentic work.
What is GPT-5.6?
GPT-5.6 is OpenAI's latest general-purpose model family for coding, professional knowledge work, computer use, scientific analysis, cybersecurity, document creation and long-running AI agents.
The number 5.6 identifies the model generation, while the names Sol, Terra and Luna identify separate capability and price tiers. OpenAI says these tiers can develop on their own release schedules instead of being treated as temporary product names.
This structure gives developers and businesses more control over cost. They can use Sol only for the hardest assignments, Terra for balanced production workloads and Luna for high-volume tasks that require lower latency and lower token costs.
GPT-5.6 Sol, Terra and Luna explained
GPT-5.6 Sol
The flagship model for complex professional assignments where accuracy, reasoning depth and persistence are the priority.
- Advanced coding and software engineering
- Long-horizon research and analysis
- Cybersecurity and vulnerability research
- Professional documents and presentations
- Large-context and multi-step agent workflows
GPT-5.6 Terra
A balanced model offering strong GPT-5.6 capabilities at half the standard Sol API token price.
- Everyday professional AI assistance
- Application and agent development
- Code generation and debugging
- Content analysis and summarisation
- Cost-sensitive business automation
GPT-5.6 Luna
The fastest and least expensive member of the family, designed for scalable and latency-sensitive workloads.
- Customer-support automation
- Classification and extraction
- High-volume content processing
- Fast application responses
- Budget-focused API deployments
GPT-5.6 Sol vs Terra vs Luna comparison
| Category | GPT-5.6 Sol | GPT-5.6 Terra | GPT-5.6 Luna |
|---|---|---|---|
| Position | Flagship model | Balanced lower-cost model | Fastest, lowest-cost model |
| Best for | Hard reasoning, coding, research and professional work | General production workloads and everyday agents | Fast, repetitive and high-volume tasks |
| Input price | $5 / 1M tokens | $2.50 / 1M tokens | $1 / 1M tokens |
| Output price | $30 / 1M tokens | $15 / 1M tokens | $6 / 1M tokens |
| Relative cost | Highest | Medium | Lowest |
| Reasoning priority | Maximum capability | Performance-cost balance | Speed and efficiency |
| API access | Available | Available | Available |
GPT-5.6 API pricing
OpenAI charges GPT-5.6 API usage according to the number of input and output tokens processed. Prices shown below are for one million tokens.
| Model | Input | Output | Example positioning |
|---|---|---|---|
| GPT-5.6 Sol | $5.00 | $30.00 | Complex and high-value professional work |
| GPT-5.6 Terra | $2.50 | $15.00 | Balanced production applications |
| GPT-5.6 Luna | $1.00 | $6.00 | Fast, high-volume and affordable workloads |
Cached-input reads receive a discount, while cache writes for GPT-5.6 and later models are billed above the normal uncached input rate. Applications using very long prompts should also review OpenAI's current long-context pricing rules before estimating their monthly cost.
Is GPT-5.6 available in ChatGPT?
Yes. OpenAI began the global GPT-5.6 rollout on July 9, 2026. Access varies according to the ChatGPT product and subscription level.
Standard ChatGPT
Plus, Pro, Business and Enterprise users can access GPT-5.6 Sol through medium and higher reasoning settings. Sol Pro is available to eligible higher-tier users for demanding work.
ChatGPT Work
Free and Go users receive access to Terra. Eligible paid plans can choose Sol, Terra or Luna and adjust the reasoning effort used for each assignment.
Codex
GPT-5.6 is available for software-development workflows, including code generation, debugging, terminal work and long-running engineering assignments.
OpenAI API
Developers can access all three models through the API and use features such as Programmatic Tool Calling and beta multi-agent coordination.
Major GPT-5.6 features
1. Max reasoning effort
The new max reasoning setting allows GPT-5.6 to spend more time exploring alternatives, checking its work and revising a solution. It is intended for difficult assignments where higher accuracy is more important than the fastest response.
2. Ultra multi-agent mode
Ultra goes beyond a single-agent workflow. OpenAI describes it as a high-capability setting that coordinates multiple agents across parallel workstreams. The default configuration uses four agents, helping complicated research, coding and analysis tasks progress simultaneously.
3. Programmatic Tool Calling
Through the Responses API, GPT-5.6 can write and run lightweight programs that coordinate tools, filter intermediate information, monitor progress and decide what action should happen next.
Instead of repeatedly sending every tool result back into the entire prompt, the model can process intermediate data in memory and retain only the information needed for the next step. This can reduce token use and the number of model round trips in complex agent workflows.
4. Better coding and frontend design
OpenAI positions GPT-5.6 Sol as its strongest coding model at launch. The model is designed to handle implementation, terminal workflows, debugging, interface creation and longer engineering assignments with less steering.
GPT-5.6 also focuses on visual quality. It can inspect rendered interfaces, refine layouts and make design corrections rather than stopping after generating the initial HTML, CSS or application code.
5. Improved documents, spreadsheets and presentations
OpenAI says GPT-5.6 can produce more polished documents, presentations and spreadsheets while following reference templates more accurately. This includes recognising layouts, typography, spacing, colours and recurring design patterns.
6. Large-context professional work
The official GPT-5.6 Sol API documentation lists a 1,050,000-token context window and a maximum output of 128,000 tokens. This is useful for analysing large codebases, extensive reports, document collections and multi-stage projects.
Official GPT-5.6 performance highlights
OpenAI published evaluation results covering coding, browsing, computer use, cybersecurity, science, long context and professional knowledge work. Benchmark results should be treated as controlled measurements rather than guarantees for every real-world prompt.
OpenAI also reported that Sol achieved a coding-agent index score of 80, while Terra and Luna reached 77.4 and 74.6 respectively. Results can vary based on reasoning level, tools, agent configuration, latency settings and evaluation methodology.
Which GPT-5.6 model should you use?
GPT-5.6 safety and limitations
More capable AI models can create both benefits and risks. OpenAI's system card classifies GPT-5.6 models as having high capability in cybersecurity and biological or chemical risk areas, but states that they did not reach the framework's highest critical threshold.
OpenAI uses layered safeguards, including protections trained into the model, real-time checks, monitoring, account-level enforcement and restricted access for particularly sensitive capabilities.
The company also reports that GPT-5.6 can sometimes go beyond the user's intended action in agentic coding tests. Therefore, businesses should not allow an AI agent to make irreversible changes without permissions, audit logs, testing environments and appropriate human review.
Frequently asked questions
When was GPT-5.6 released?
OpenAI launched the GPT-5.6 family for general availability on July 9, 2026, following an earlier limited preview.
What are GPT-5.6 Sol, Terra and Luna?
They are three GPT-5.6 capability tiers. Sol is the flagship model, Terra provides a balance between performance and cost, and Luna is the fastest and most affordable option.
Is GPT-5.6 available to free ChatGPT users?
OpenAI states that Free and Go users can access GPT-5.6 Terra through ChatGPT Work. Access to Sol and other settings depends on the product and subscription plan.
How much does the GPT-5.6 API cost?
Per one million tokens, Sol costs $5 for input and $30 for output, Terra costs $2.50 for input and $15 for output, and Luna costs $1 for input and $6 for output.
What is GPT-5.6 Ultra mode?
Ultra is a high-capability setting that coordinates multiple agents across parallel workstreams. OpenAI's default Ultra configuration uses four agents.
Which GPT-5.6 model is best for coding?
Sol is OpenAI's flagship choice for the hardest coding and software-engineering tasks. Terra may offer better value for many everyday development workloads, while Luna is suitable for faster and lower-cost code-related tasks.
Does GPT-5.6 support images?
The GPT-5.6 Sol API documentation lists text and image input, with text output. Availability of particular modalities may vary between models, endpoints and products.
Will GPT-5.6 always be more accurate?
No AI model is correct in every situation. Performance depends on the prompt, source material, tools, reasoning setting, workflow design and the type of task being completed.
Final verdict
GPT-5.6 represents more than a single model upgrade. The Sol, Terra and Luna structure allows individuals, developers and organisations to choose different levels of capability, speed and cost for different workloads.
Sol is the best fit for difficult, high-value assignments. Terra offers a practical balance for everyday production use. Luna provides the lowest token price and fastest positioning for large-scale applications.
The best model should be selected using real task-completion cost, accuracy, latency and human-review requirements—not model size or benchmark scores alone.