GPT-6 Sol and GPT-6 Luna
GPT-6 Sol and Luna bring Astra-era training to cheaper tiers at half the price of their GPT-5.6 equivalents.
Sol at $2/$10 and Luna at $0.10/$0.50 per million tokens, 1M-token context. On DeepSWE v1.1 Sol at max effort scores 68.8%, about 1.1 points under Claude Fable 5 at roughly 80% lower cost per task. Luna is among OpenAI's cheapest models ever; GPT-5.6 is scheduled for a 25% price rise in November.
- Date
- Tuesday, 22 September 2026
- Lab
- OpenAI
- Kind
- model
- Access
- closed API
- Price
- Sol $2/$10; Luna $0.10/$0.50 per 1M tokens (2026-09-22)
Figures
| Measure | Value | Measured by |
|---|---|---|
| DeepSWE v1.1 (Sol, max effort) | 68.8% Claude Fable 5 69.9% at xhigh; about 80% lower cost per task | company |
| API price Sol / Luna (input per 1M tokens) | $2 / $0.10 output $10 / $0.50; down from $4 / $0.20 | company |
Launched about an hour after Anthropic's Claude Opus 5.5 (Simon Willison); he calls it a new price war. Not yet in ChatGPT Chat at launch.
Sources
- openai.com/index/introducing-gpt-6-sol-and-luna/
- simonwillison.net/2026/Sep/22/opus-and-sol-and-luna/
- developers.openai.com/api/docs/changelog
This record was checked against its sources on 6 October 2026. How we check