Rahul Somvanshi
If you pay for AI tools, whether as a developer, a business or someone with a ChatGPT subscription, the cost of getting work done just dropped again. On Tuesday, OpenAI released GPT-6 Sol and GPT-6 Luna at half the API price of the models they replace. The launch came about 90 minutes after Anthropic released its own cheaper model, Claude Opus 5.5. Two of the biggest AI companies cut prices on the same afternoon.
OpenAI launched GPT-6 Sol and GPT-6 Luna on September 22, 2026. That is about three weeks after its flagship model, GPT-6 Astra, arrived earlier this month. Sol is the mid-range model, built for coding and multi-step work. Luna is the smallest and cheapest in the GPT-6 family.
API prices are $2 per million input tokens and $10 per million output tokens for Sol. Luna costs $0.10 and $0.50. A token is a small chunk of text, roughly three-quarters of a word. OpenAI says these prices are 50% lower than the promotional pricing for the earlier GPT-5.6 Sol and Luna.
The models are rolling out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. They are also available through the API.
For ChatGPT subscribers, the change shows up as better model options with no price rise. For businesses and developers paying by usage, it cuts the bill for the same work. A company spending $1,000 a month on the older Sol through the API would pay about half that for the same volume on GPT-6 Sol, if OpenAI’s 50% figure holds for its workload. Free and Go users can try GPT-6 Luna in the ChatGPT desktop app.
What GPT-6 Sol and Luna do differently
OpenAI’s headline claim is about accuracy. “On our internal factuality evaluation, GPT-6 Sol makes about half as many mistakes as its predecessor, approaching Astra-level reliability at much lower cost,” the company said.
In plain terms, Sol is meant to give answers close to the flagship model’s quality without the flagship price.
Luna is aimed at high-volume, simple jobs. OpenAI lists summarising documents, pulling out information and answering quick questions. The company says Luna, at higher effort settings, “matches GPT-5.6 Sol at about a hundredth its cost.”
OpenAI puts the lower prices down to engineering. “Improvements in caching and inference let us serve these models at lower cost, and we’re passing those savings directly on to users and customers,” it said. Caching means the system reuses work it has already done on repeated text, much like a browser storing a page so it loads faster the second time.
GPT-6 price war: why Anthropic moved first
Anthropic released Claude Opus 5.5 about 90 minutes before OpenAI’s announcement, according to TechCrunch. Anthropic cut Opus 5.5’s list price by 20% to $4 per million input tokens and $20 per million output tokens. It says the model costs 40% less to run than Opus 5 on typical workloads. We cover the details in our Claude Opus 5.5 breakdown.
Both companies are making the same pitch: more capable models that cost less per task. OpenAI made that case in its GPT-5.6 announcements earlier this year, and Anthropic’s Opus 5.5 launch leans on it too.
The companies’ own benchmark claims are hard to compare directly. OpenAI told TechCrunch that GPT-6 Sol and Luna “handle tasks substantially better than Anthropic’s top models.” Anthropic published its own scores showing gains for Opus 5.5. Each company chooses its own tests and settings, so the numbers do not line up like for like.
Readers can follow earlier coverage on our OpenAI and artificial intelligence pages.
Where developers can use GPT-6 Sol and Luna
GitHub says both models are coming to GitHub Copilot. GPT-6 Sol is available on Copilot Pro+, Max, Business and Enterprise plans. GPT-6 Luna is available on those plans and on Copilot Pro. Both can be picked in VS Code, Visual Studio, JetBrains, Xcode, Eclipse, the Copilot CLI and GitHub Mobile. “Rollout will be gradual. Check back soon if you don’t see the models yet,” GitHub said.
Business and Enterprise administrators control access. New models switch on by default unless an admin has turned them off.
The pricing applies wherever developers pay OpenAI in US dollars. That covers the US, UK and Australia alike. The ChatGPT app rollout follows each subscription tier, not the user’s country.
So what is GPT-6 Sol? It is OpenAI’s mid-tier GPT-6 model for coding and multi-step work, priced at $2 per million input tokens and $10 per million output tokens. OpenAI says it makes about half as many factual mistakes as GPT-5.6 Sol. It is available now in ChatGPT, Codex, the API and GitHub Copilot.
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22 at half the API price of their GPT-5.6 versions. Sol targets coding and complex tasks. Luna handles quick, high-volume jobs. The launch came 90 minutes after Anthropic’s cheaper Claude Opus 5.5. Rollout across ChatGPT and GitHub Copilot is gradual. Anthropic says Claude Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks. Check back for pricing comparisons when they land.
—
