OpenAI Just Slashed API Prices in Half — Here’s What GPT-6 Sol and Luna Mean for Your Business
OpenAI released two new models yesterday: GPT-6 Sol and GPT-6 Luna. They’re not the company’s most powerful options — that title still belongs to GPT-6 Astra — but they’re positioned to be the workhorses that most businesses will actually use. And the pricing is genuinely significant.
Luna starts at $0.10 per million input tokens and $0.50 per million output tokens. Sol sits in the middle at $2.00 input and $10.00 output. Compare that to Astra at $10/$50, and you start to see what OpenAI is building: a three-tier system where you pick the right tool for the job rather than paying top dollar for everything.
That’s a smart move — and a direct response to pressure from Anthropic and Google, both of whom have been competing hard on price.
What the Tiers Actually Mean
Think of it like cloud compute instance types. You don’t run a database on your most expensive VM when a mid-range one handles it just fine.
Luna is for high-volume, lower-complexity tasks: summarizing documents, classifying support tickets, extracting data from forms, generating first drafts. Fast and cheap. If you’re running thousands of these per day, Luna cuts your bill dramatically.
Sol is for work that requires more nuance — writing code, analyzing contracts, reasoning through business logic, generating content that actually sounds human. It’s 50% cheaper than the GPT-5.6 models it effectively replaces, while delivering meaningfully better results.
Astra stays in the mix for genuinely hard problems: complex multi-step reasoning, deep research tasks, anything where accuracy matters more than cost.
Why This Changes the Calculation for AI Automation
One of the most common things we hear from businesses exploring AI automation is: “The demos looked great, but the API costs don’t pencil out at scale.”
That calculation just shifted. A workflow that was costing $500/month in API calls using previous-generation models might now run on Luna for under $50. That’s not a minor adjustment — that’s the difference between a pilot that gets shelved and a production system that pays for itself.
For businesses that built automations on GPT-4 or GPT-5-era models and never revisited the infrastructure: now’s a good time to audit. Many of those workflows were over-engineered for the model that was available at the time. Luna might handle 80% of them just fine.
One Caveat Worth Knowing
Cheaper models aren’t free models. Luna and Sol still hallucinate. They still need well-structured prompts, clear instructions, and human review on anything that matters. The price drop doesn’t eliminate the need for thoughtful implementation — it just makes poor implementations more affordable to run badly at scale.
The businesses that get real value from this will be the ones who design their workflows carefully: Luna for triage, Sol for reasoning, Astra only when genuinely needed.
The Practical Bottom Line
If you’ve been waiting for AI automation to become cost-effective enough to justify, the window just opened wider. The question isn’t whether these tools are ready — it’s whether your implementation plan is.
Want to explore how the new GPT-6 model tiers could fit into your business workflows? Let’s talk.