OpenAI releases GPT-6 Sol and Luna at half the GPT-5.6 promotional price
On 22 September, the same day as Claude Opus 5.5, OpenAI added two cheaper GPT-6 models. Sol and Luna are meant to carry Astra's gains into everyday work. API prices are half the GPT-5.6 promotional rates. They are not a replacement for Astra on the hardest jobs, by OpenAI's own account.
Published · 1 source
Confirmed
Confirmed. The models, the 50% price cut against GPT-5.6 promotional rates, and the day-one surfaces are in OpenAI's own post. A pricing-page change, or a wider ChatGPT rollout note, would be what would move this entry.
Why this matters
If you pay the API, the numbers are simple. GPT-6 Sol is $2 per million input tokens and $10 per million output tokens, down from $4 and $20 for GPT-5.6 Sol's promotional price. GPT-6 Luna is $0.10 and $0.50, down from $0.20 and $1.20. OpenAI says caching and inference got cheaper and that it is passing that on.
Access on day one is ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu, plus the API. OpenAI still points the most demanding work at GPT-6 Astra.
What we know
- OpenAI released GPT-6 Sol and GPT-6 Luna on 22 September 2026, describing them as a way to spread GPT-6 Astra's gains across everyday work at lower cost. [1]
- API prices fell 50% versus GPT-5.6 promotional pricing: Sol is $2 per million input tokens and $10 per million output tokens; Luna is $0.10 and $0.50. [1]
- Both models rolled out the same day in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, and in the API. OpenAI says the most demanding projects still call for Astra. [1]
What is unclear
- Which ChatGPT plans outside Work and Codex can select Sol or Luna, and when. The announcement names those surfaces and the paid tiers above.
- How the quality compares with Astra on a task you care about. OpenAI describes a cost move, not a new top of the range.
Sources
- Official announcement
OpenAI
openai.com
Primary source
Open original (opens in a new tab)
Learn the idea behind this story
Sol and Luna are a direct lesson in matching the model to the job: cheaper where the work is routine, Astra where it is not.
Open the prompt lab