GPT-6 Sol and Luna: What Do the New Models Offer Businesses?
GPT-6 Sol and Luna models are now live. We break down their pricing, performance, and which model fits which task.
Contents 6
OpenAI released GPT-6 Sol and Luna on September 22, 2026. Sol is a balanced model built for complex, agentic coding tasks, while Luna is designed for fast, low-cost jobs. Together, these two models make it easier for businesses to pick the right tool for the type of task at hand. This comparison clarifies which model fits which job.
In short:
- GPT-6 Sol and Luna launched on September 22, 2026.
- Sol is optimized for agentic coding; Luna is optimized for speed and low cost.
- Sol costs $2/$10 per million tokens, while Luna costs $0.10/$0.50.
- Both models share a 1.05M-token context window.
What are GPT-6 Sol and Luna?
These two models are the latest releases following GPT-6 Astra. They were trained using methods similar to Astra (OpenAI). The goal is to make progress in professional work and coding faster and more affordable. This approach gives teams with different budget and speed needs more options to choose from.
Sol is positioned as a balanced model for interactive and agentic coding. Luna is the most affordable option in the GPT-6 family. Both are available within GitHub Copilot at different plan tiers.
GPT-6 Sol vs Luna: which one should I choose?
The choice depends on task complexity and your budget. Sol stands out in complex coding tasks that require more reasoning ability (MacRumors). Luna is optimized for fast responses and high-volume operations.
For simple automation tasks, rapid prototyping, or systems that send a large number of requests, Luna is a sensible choice. For complex architectural decisions, multi-step agent tasks, or work that demands high accuracy, Sol is the better fit. Businesses often use both models together to balance cost and performance.
How does the cost comparison look?
Pricing is the most obvious difference between the two models. Sol costs $2 per million input tokens and $10 per million output tokens (The New Stack). Luna is far cheaper: $0.10 for input and $0.50 for output.
That gap translates into meaningful savings for high-volume workloads. At the high-effort setting, Luna shows a 5.4-percentage-point improvement over the previous model while costing 58 percent less per task (OpenAI). This shows it's now possible to achieve the same quality at a much lower cost.
| Feature | GPT-6 Sol | GPT-6 Luna |
|---|---|---|
| Input cost | $2 / million tokens | $0.10 / million tokens |
| Output cost | $10 / million tokens | $0.50 / million tokens |
| Knowledge cutoff | April 20, 2026 | May 18, 2026 |
| Context window | 1.05M tokens | 1.05M tokens |
| Max output | 128K tokens | 128K tokens |
How much of an improvement is this over previous models?
GPT-6 Sol makes roughly half as many errors as GPT-5.6 Sol (Jake Handy). That's a significant step forward for coding tasks that demand accuracy. Luna, meanwhile, in high-effort mode reaches results comparable to GPT-5.6 Sol at roughly one percent of the cost.
These gains show that the improvements that began with GPT-6 Astra are now trickling down to cheaper tiers. Advances in accuracy, coding, and computer use are now reaching a much broader user base. This is notable when compared with the similar competitive dynamics seen with models like Claude Opus 5.5.
What are the technical specifications?
Both models share a 1.05 million token context window (Kingy AI). Maximum output is capped at 128K tokens. Both support text and image input.
The models can also use OpenAI's core agentic tools, including function calling, web search, file search, and computer use. This toolset is a significant advantage for developers looking to build agentic workflows. Both models support these tools in a similar way.
Which plans can I use them with?
GPT-6 Sol is available on Copilot Pro+, Max, Business, and Enterprise plans (GitHub Changelog). Luna has broader availability, accessible across Pro, Pro+, Max, Business, and Enterprise plans.
This distribution shows that Luna is also accessible to entry-level users. Sol, on the other hand, is limited to higher-tier plans. Businesses can choose their plan based on team size accordingly.
When will these be available in ChatGPT?
These models are currently accessible only through GitHub Copilot. They are not yet available in ChatGPT's chat mode (MacRumors). It's still unclear whether OpenAI will bring these models to its other products in the future.
This suggests the company is prioritizing the developer and coding ecosystem for now. The GitHub Copilot integration stands out as part of that strategy.
Frequently asked questions
What is the key difference between GPT-6 Sol and Luna?
Sol offers more reasoning power for complex, agentic coding tasks. Luna is focused on speed and low cost. You can choose between the two based on task complexity.
How current is the knowledge in these models?
Sol's knowledge cutoff is April 20, 2026, while Luna's is May 18, 2026. These dates indicate when each model's training data ends.
How big is the cost advantage?
Luna is far cheaper than Sol in both input and output token costs. It also costs 58 percent less per task compared to the previous model. That's a significant saving for high-volume use.
Where can I use these models?
For now, they're only accessible through GitHub Copilot. Sol is available on Pro+, Max, Business, and Enterprise plans, while Luna is available on all plans, including Pro. No date has been announced yet for ChatGPT integration.
These two models are reshaping the cost-performance balance for coding and agentic tasks. The result is two strong options suited to different budgets. Choosing the right model for a business depends on the complexity of the project and the available budget. At EngerekTech, we help teams that want to integrate this kind of AI tooling into their software processes.
Sources
