GPT-6 Sol vs GPT-6 Luna: OpenAI's New Tiers Explained

By ChatGBot · September 29, 2026

On September 22, 2026, about ninety minutes after Anthropic shipped Claude Opus 5.5, OpenAI released two new GPT-6 tiers: GPT-6 Sol and GPT-6 Luna. They are not about topping the leaderboard. They are about cost per task, and Luna in particular resets how cheap a capable model can be. Here is the difference and which one to use.

GPT-6 Sol: the workhorse for hard tasks

Sol is priced at $2 per million input tokens and $10 per million output tokens, a tier below the top Astra model. It is tuned for heavier work like coding, multi step reasoning, and tasks where quality matters more than raw volume. If you are shipping code or solving something genuinely hard, Sol is the one to reach for. New to the family? Start with what is GPT-4o.

GPT-6 Luna: a new price floor

Luna is the story of the launch. At $0.10 per million input tokens and $0.50 per million output tokens, with cached input at $0.01, it sets a new mainstream price floor and undercuts even DeepSeek's fast tier. It is built for high volume work: summarizing documents, extracting data, classifying, and anything you run thousands of times where cost adds up fast.

Which one should you use?

• Pick Sol for coding, analysis, and hard reasoning where a wrong answer is expensive.
• Pick Luna for summaries, extraction, tagging, and any high volume job where speed and cost matter most.
• Use both and let the task decide. A frontier model for the hard 10%, a cheap model for the routine 90%, is how the cost math actually works out.

The bigger picture: the AI price war

Sol and Luna landed the same day as Claude Opus 5.5, and within weeks of new releases from DeepSeek, Qwen and GLM. Prices are falling across the board, and the real advantage now goes to whoever can match the right model to each job. For the full rundown, read our September 2026 AI model releases roundup.

Frequently asked questions

What is the difference between GPT-6 Sol and Luna? Sol ($2 / $10 per million tokens) is tuned for complex work like coding, while Luna ($0.10 / $0.50) is a low cost tier for high volume tasks.

Is GPT-6 Luna the cheapest model? At $0.10 per million input tokens, Luna set a new mainstream price floor in September 2026, undercutting even fast tiers from other labs.

When were GPT-6 Sol and Luna released? September 22, 2026, about ninety minutes after Anthropic released Claude Opus 5.5.

Can I use GPT-6 in ChatGBot? ChatGBot gives you GPT along with Claude, DeepSeek, Qwen and GLM under one subscription, so you can pick the right model for each task.