On September 22, OpenAI released updated versions of its Sol and Luna working models, part of the GPT-6 family, available through the API and to ChatGPT Work and Codex users. The company says that, compared to the previous generation (5.6), the new models cost half as much through the API — a reduction it attributes to improvements in caching and in how requests are processed (inference). Separately, OpenAI says Sol makes half as many factual mistakes as its predecessor, approaching Astra-level reliability; no such figure is given for Luna. Sol is built for technical work and coding, Luna for administrative tasks such as summaries, pulling information out of documents, and answering straightforward questions.
Why it matters for a small business
The cost per query is the main reason many small companies still don't use AI at scale, beyond the occasional test run. The API price cut matters directly if you build or use tools running on top of OpenAI's API — chatbots, automations, integrations — for repetitive tasks like sorting messages or drafting a first reply to a client, a lower cost per request changes the math.
If you only use the ChatGPT interface rather than the API directly, check with your provider or in your account settings whether this price change applies to you — the source is about the API, not the chat subscription.
Source: TechCrunch
