In plain English
- What happened
- OpenAI released GPT-6 Sol and GPT-6 Luna, two lower-cost models built with the same methods as its flagship GPT-6 Astra. API prices for both dropped 50% compared to their GPT-5.6 promotional rates, and the models are now live in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu accounts.
- Why it matters to your business
- Most day-to-day business tasks — drafting an email, summarizing a call, answering a customer question — do not need the most expensive model available. Sol and Luna give ChatGPT subscribers and API users a cheaper option for that everyday work, which matters if you or a vendor tool is paying per token.
- Do this
- Open your ChatGPT model picker and check whether Sol or Luna now appears as an option for routine tasks, so you are not paying for Astra-level power you do not need.
What OpenAI released
Earlier in the same month, OpenAI introduced GPT-6 Astra as its most capable model. Sol and Luna are smaller, cheaper models trained with similar methods, meant for work that does not require Astra's full depth. OpenAI describes them as advancing the frontier on cost efficiency rather than raw capability.
The company says it improved caching and inference so it could serve these models more cheaply, and is passing that savings on directly through lower API prices.
What it costs
For GPT-6 Sol, input token price dropped from $4 to $2 per million tokens, and output from $20 to $10. For GPT-6 Luna, input dropped from $0.20 to $0.10 and output from $1.20 to $0.50 per million tokens. Both are 50% cuts from the previous GPT-5.6 promotional pricing.
| Model | Input | Output | Change |
|---|---|---|---|
| GPT-6 Luna | $0.20 → $0.10 | $1.20 → $0.50 | 50% cheaper |
| GPT-6 Sol | $4 → $2 | $20 → $10 | 50% cheaper |
| GPT-6 Astra | $10 | $50 | unchanged |
API input and output prices for both new models are 50% lower than their GPT-5.6 predecessors, per OpenAI's own pricing table.
Who can use it, and where
GPT-6 Sol and Luna are available now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu subscribers. Free and Go users can access Luna specifically through the desktop app. Neither model is yet available in the standard Chat interface.
In the API, developers can call the models directly as gpt-6-sol and gpt-6-luna. OpenAI says it is rolling the models out gradually through the day to keep the service stable, so a subscriber who does not see the new options right away may need to check back later.
What changed besides price
OpenAI says both models carry over Astra's alignment work, including lower rates of misleading claims about coding work in its internal evaluations, and a communication style with less jargon and fewer unnecessary details. It also reports internal factuality tests showing Sol makes about half as many mistakes as its predecessor.
For developers building automated workflows or chatbots on the API, OpenAI also improved prompt caching, which reuses parts of a conversation instead of reprocessing them, cutting costs further for repeated or long-running tasks.
Reality check
OpenAI's benchmark comparisons against competitor models come from its own internal tests and evaluation environment, which can behave differently from the production ChatGPT app. If cost matters to your workflow, test the cheaper model on your own real tasks before switching over, rather than trusting a benchmark score alone.
Sources
