GPT-6 Sol and Luna: Performance, Pricing & Key Features

The most important thing is not just finding the most powerful model. For developers and businesses, cost, speed, reliability, coding ability, and how well a model handles long workflows can be just as important.

That’s where GPT-6 Sol and Luna can help. After the release of GPT-6 Astra, OpenAI has introduced Sol and Luna as more affordable options for professional work, coding, computer use, and high-volume AI applications. The goal is to make advanced AI more practical for different budgets and workloads.

Key Takeaways

• GPT-6 Sol and Luna bring improvements in professional work, coding, factuality, computer use, communication, and model efficiency.
• GPT-6 Sol costs $2 per 1M input tokens and $10 per 1M output tokens, while GPT-6 Luna costs $0.10 input and $0.50 output per 1M tokens.
• GPT-6 Sol delivers stronger performance for complex coding, professional workflows, and computer-use tasks, while GPT-6 Luna focuses on lower-cost performance.
• Both models introduce improved prompt caching, with cached input reads receiving a 90% discount, helping reduce costs for repeated context.
• GPT-6 Sol and Luna are available through selected ChatGPT Work, Codex, and API access, with availability depending on the plan and rollout stage.

What Are GPT-6 Sol and GPT-6 Luna?

There are two models in OpenAI’s GPT-6 family: GPT-6 Sol and GPT-6 Luna. The GPT-6 Astra is the model designed for the most demanding work, while the Sol and Luna are intended to provide strong capabilities at a lower cost.

Sol focuses more on complex professional tasks, coding, and agentic workflows. Luna is designed to be efficient and handle a lot of requests, which makes it a cost-effective option for developers.

GPT-6 Sol and Luna Pricing and Features

One of the biggest changes is the reduction in API pricing. According to OpenAI’s official announcement, GPT-6 Sol and Luna are priced 50% lower than the promotional pricing of their GPT-5.6 counterparts.

The published API prices per 1 million tokens are:

• GPT-6 Sol: $2 input / $10 output
• GPT-6 Luna: $0.10 input / $0.50 output
• GPT-5.6 Sol: $4 input / $20 output
• GPT-5.6 Luna: $0.20 input / $1.20 output

This makes the GPT-6 Sol and Luna pricing and features particularly relevant for developers building AI products where API costs can increase quickly with usage. The actual cost of an application will still depend on token consumption, caching, reasoning effort, and the type of workload.

GPT-6 Sol: Key Features

GPT-6 Sol is for developers who need better performance on complex tasks, but don’t always need the top-of-the-line Astra model. It can handle professional workflows, coding, computer-use tasks, and other applications where multiple steps and deeper reasoning are required.

The model can also benefit from improved caching, which is a way to store data so that applications that reuse context can reduce processing costs and improve efficiency. Sol is especially useful for coding agents and business applications that need to find a balance between what it can do and how much it costs.

Key areas include:

• Complex coding
• Professional workflows
• AI agents
• Computer use
• Research and analysis
• Long-running tasks
• Tool-based workflows

GPT-6 Luna: Key Features

GPT-6 Luna is designed to be more efficient. It is intended for developers that need capable AI across a large number of requests while keeping the cost per task low.

Luna can be useful for tasks like processing content, automating processes, and creating summaries. Its lower API cost can become especially important when an application handles thousands or millions of requests.

Common use cases include:

• High-volume automation
• Content processing
• Summarization
• Data extraction
• Classification
• AI-powered applications
• Repetitive workflows

GPT-6 Sol API Pricing and Performance

The GPT-6 Sol API is priced and performs well, making it a good option for teams building more demanding AI applications. Sol is considerably cheaper than GPT-5.6 Sol’s previous promotional rates. It costs $2 per million input tokens and $10 per million output tokens.

OpenAI’s published evaluations show improvements across several professional and technical benchmarks. For example, on AutomationBench, GPT-6 Sol at xhigh effort scored 33.2% at a reported cost of $0.27 per task. These results can help you compare different options, but the performance in the real world can be different depending on the application’s prompts, tools, and workflow.

GPT-6 Luna Features and API Pricing

The GPT-6 Luna is a great option for applications that require cost efficiency. The price is $0.10 for one million input tokens and $0.50 for one million output tokens.

OpenAI also says that its performance has improved compared to GPT-5.6 Luna in several tests. This combination of lower pricing and improved capability could be useful for products that need to process large amounts of information created by AI or analyzed using AI.

Performance Across Professional Work

AI agents are being used more and more for tasks that involve many applications and several steps that are connected to each other. GPT-6 Sol is designed to handle this type of professional workflow while costing less than higher-priced models.

On AutomationBench, OpenAI reports that GPT-6 Sol at high effort achieved a 33.2% score at $0.27 per task. In the same evaluation, GPT-6 Luna did 5.4 percentage points better than the previous version, but cost 58% less per task.

Factuality and Reliability

It’s not enough to get an answer quickly if the information has mistakes. OpenAI says GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol in its internal factuality test, which makes its performance more like GPT-6 Astra at a lower cost.

GPT-6 Luna also shows improvements at higher reasoning levels. OpenAI says that Luna can match GPT-5.6 Sol’s performance on this test for a lower cost. But these results come from a specific internal test, so they may not be the same for every real-world task.

Coding Performance

Coding is another major area where OpenAI has focused on improving GPT-6 Sol and Luna. Developers are using AI agents more and more for large software projects. This makes it important to have good models and not spend too much money on long coding sessions.

OpenAI says that GPT-6 Sol performs much better than GPT-5.6 Sol in the FrontierCode evaluation. On the DeepSWE test, GPT-6 Sol performed at its best and scored 68.8%, while GPT-6 Luna scored 66.6%. This shows that both models can handle difficult software engineering tasks.

Computer Use

People who work with computers need to understand how to use different interfaces, be able to do many things at once, and recover from unexpected situations. GPT-6 Sol and Luna are designed to improve this type of workflow while keeping costs below those of more expensive models.

In the offline evaluation of OSWorld 2.0, OpenAI reported a score of 60.5% for GPT-6 Sol at high effort, compared to 60.3% for Claude Opus 5 at medium effort. GPT-6 Luna performed better than GPT-5.6 Sol at both high and medium levels of effort in the evaluation.

GPT-6 Sol vs GPT-5.6 Sol

The GPT-6 Sol is better than the GPT-5.6 Sol in a number of ways. It has more capabilities and is more affordable. GPT-6 Sol costs half as much as GPT-5.6 Sol’s promotional price, and OpenAI has seen better results in several coding and professional work tests.

The models should still be compared according to the actual requirements of an application. If you’re a developer moving from GPT-5.6 Sol, it’s important to test things like representative prompts, tool calls, latency, output quality, and total cost before changing your production workflow.

GPT-6 Luna vs GPT-5.6 Luna

The GPT-6 Luna is compared to the GPT-5.6 Luna in terms of efficiency and improved performance. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. This is compared with $0.20 and $1.20 for GPT-5.6 Luna’s promotional pricing.

OpenAI says that Luna is better than its predecessor in several ways, while still being low-cost. For businesses that process a lot of requests, even small differences in cost per million tokens can have a big impact on overall API spending.

Improved Prompt Caching

GPT-6 has more than one way to reduce costs, and lower token prices is just one of them. OpenAI has also improved prompt caching for GPT-6 models. This helps applications reuse context instead of repeatedly processing the same information.

OpenAI says that cached input tokens get a 90% discount. Developers can also use the Prompt Caching Dashboard and diagnostics tools to understand cache performance and identify opportunities to improve reuse.

Better Communication Style

OpenAI says it has also brought some of GPT-6 Astra’s communication improvements to Sol and Luna. The goal is to make responses clearer, use less jargon, and add only details that help answer the question.

This is especially helpful in technical conversations. Instead of explaining things in great detail, the models just communicate the important information in a clear way.

Alignment and Safety Improvements

GPT-6 Sol and Luna build on the alignment work introduced with GPT-6 Astra. OpenAI says that its new version is better than the GPT-5.6 versions in a number of internal safety tests, including tests that checked if claims about coding work were true.

These evaluations are intentionally designed to test difficult scenarios and aren’t meant to represent normal user behavior. Developers should therefore consider the published safety evaluations alongside their own testing and monitoring when deploying AI agents.

What These Models Mean for Developers

The main change is that developers have more options at different levels of capability and cost. Instead of using the same expensive model for every request, teams can choose a model based on how complicated the task is and how often it needs to run.

For example, a business might use Sol for complex coding or agentic workflows while using Luna for high-volume classification or content-processing tasks. This type of model selection can help teams manage performance requirements and operating costs.

GPT-6 Sol and Luna Availability

You can find GPT-6 Sol and GPT-6 Luna through the OpenAI API. Their codes are gpt-6-sol and gpt-6-luna. OpenAI has also made the models available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users.

If you use Free or Go, you can use GPT-6 Luna in the desktop app. At the time of the announcement, the models weren’t available in the regular Chat experience. They were gradually being added to ChatGPT Work and Codex.

Final Thoughts

GPT-6 Sol and Luna give developers more options when building AI-powered products and workflows. Sol is better for complex professional work and coding, while Luna is better for applications that have a lot of traffic.

The biggest changes are not only about benchmark scores. Lower API pricing, better caching, stronger coding performance, and better efficiency can affect the practical cost of running AI applications. Developers should test these models using their own workloads instead of using only one benchmark.

If your team is thinking about upgrading, it’s best to compare how well the tools work, how long they take to respond, how easy they are to use, how much data they use, how reliable they are, and how much they cost for each task. This makes it easier to see if GPT-6 Sol or GPT-6 Luna is better for a specific task.

FAQs

1. What are GPT-6 Sol and GPT-6 Luna?

GPT-6 Sol and GPT-6 Luna are newer OpenAI models designed to improve professional work, coding, factuality, computer use, and overall efficiency. Sol is positioned for more demanding tasks, while Luna focuses on delivering useful performance at lower cost.

2. How much do GPT-6 Sol and GPT-6 Luna cost?

According to OpenAI’s announcement, GPT-6 Sol costs $2 per 1 million input tokens and $10 per 1 million output tokens. GPT-6 Luna costs $0.10 per 1 million input tokens and $0.50 per 1 million output tokens.

3. What are the main features of GPT-6 Sol?

GPT-6 Sol focuses on professional workflows, coding, factual accuracy, computer-use tasks, and improved collaboration. It also benefits from improved prompt caching, which can reduce the cost and processing time of repeated context.

4. What is the difference between GPT-6 Sol and GPT-6 Luna?

The main difference is their performance and cost positioning. GPT-6 Sol is designed for more complex professional and technical workloads, while GPT-6 Luna provides a faster and more cost-efficient option for many everyday and developer tasks.

5. Are GPT-6 Sol and GPT-6 Luna available in ChatGPT and the API?

OpenAI announced GPT-6 Sol and GPT-6 Luna for ChatGPT Work and Codex for eligible paid plans, with GPT-6 Luna also available to Free and Go users in the desktop app. For developers, the announced API model IDs are gpt-6-sol and gpt-6-luna. Availability may change as OpenAI continues the rollout.