OpenAI cuts GPT-6 Sol, Luna prices 50% in ChatGPT Work
OpenAI cut GPT-6 Sol and Luna prices 50% and put them in ChatGPT Work and Codex today, with 90% caching discounts by default.
OpenAI put two new models, GPT-6 Sol and GPT-6 Luna, into ChatGPT Work and Codex on September 22, cutting API prices on both by 50% from their prior promotional rates. Sol drops to $2 per million input tokens and $10 per million output tokens. Luna drops to $0.10 and $0.50, live today for most plans.
What shipped, and when
OpenAI's post, published September 22, 2026, describes Sol and Luna as the cost-efficient half of the GPT-6 family, sitting under GPT-6 Astra, which launched earlier in the month. Sol's API rate fell from $4 to $2 per million input tokens and $20 to $10 per million output tokens. Luna fell from $0.20 to $0.10 input and $1.20 to $0.50 output. OpenAI ties the cut to improvements in caching and inference that lowered its own cost to serve the models.
Availability lands the same day: Sol and Luna are in ChatGPT Work and Codex today for Plus, Pro, Business, Enterprise and Edu accounts. Free and Go users get Luna in the desktop app. Neither model is in standard ChatGPT Chat yet, and the rollout inside Work and Codex is staged through the day rather than instant for every account.
Why an owner should care
A price cut on a model you were not using is nobody's news. A price cut on the model your team runs proposals and scheduling through is different: it changes what you can afford to ask the tool to do, more drafts, more revisions, more runs before the bill notices.
The caching change matters just as much for a recurring workflow. A 90% discount on cached input by default means a tool that reuses the same template, a proposal or an intake form, gets cheaper every time it runs that pattern again.
What it looks like inside a shop
An AV or trades owner using ChatGPT Work for scopes of work and vendor emails gets the same tool at half the per-token cost today, no new sign-up beyond picking the model. A shop running Codex for small internal tools, a job tracker, a quoting helper, gets more room to iterate before hitting a usage ceiling.
Any service business runs the same math. A landscaper drafting estimates, a contractor tracking change orders, a design studio managing revisions: all of it gets cheaper the moment Sol or Luna replaces a pricier default model.
What to do this week
- If on ChatGPT Work, Plus, Pro, Business, Enterprise or Edu, open the model picker today and select GPT-6 Sol or Luna for a task you run often, a proposal draft, a scope of work, a client email.
- If running Codex for any internal tool, switch its default model to Sol or Luna and compare output quality against whatever it was using before.
- If building anything on the API, check the new caching dashboard to see how much of your usage is already hitting cache, and where it is not.
Questions owners ask
Do I need to do anything to get the lower price?
No. The lower price is already live, just select Sol or Luna where the model picker is available.
Is this in regular ChatGPT, the one most people use day to day?
Not yet. Sol and Luna are in ChatGPT Work and Codex only for now, with Free and Go users getting Luna in the desktop app. Standard ChatGPT Chat is not included.
Does the price cut affect quality?
OpenAI's own benchmarks describe both models as improving on their predecessors, not just getting cheaper. Test that against your own documents before you rely on it. Every service business runs on the same underlying leak: work that only moves when the owner is the one doing it. See how AI systems close that gap for AV and smart home companies, or read how ChatGPT drafts inside Microsoft Word for a related workflow already live. Source: Introducing GPT-6 Sol and Luna, OpenAI, September 22, 2026
