OpenAI says the updated models cut factual errors and coding mistakes while lowering API costs through better caching and inference.
OpenAI released updated versions of its GPT-6 Sol and Luna models, roughly a month after launching GPT-6 Astra. The company says the new models match Astra-level factual reliability at a lower cost, and it cut API pricing to half of the GPT-5.6 series of Sol and Luna.
Sol and Luna first appeared earlier this year as different tiers in OpenAI’s lineup. Sol targets complex work such as coding, while Luna handles high-volume clerical tasks like summarizing documents, extracting information, and answering quick questions. OpenAI also competes directly with Anthropic, whose top models include Fable and Opus.
The price cut matters most for teams running large volumes of API calls. Cheaper caching and inference bring the per-request cost down, which changes the math on document processing, data extraction, and support automation. OpenAI also claims fewer coding errors, so developers can lean on Sol for more production work.
The new Sol and Luna versions are live in ChatGPT Work, Codex, and the ChatGPT API for most paid accounts. Luna also reaches the desktop app and Free and Go users, and OpenAI plans to roll both models out across the ChatGPT app and website gradually during the day.
What matters
- OpenAI released updated GPT-6 Sol and Luna models that it says make half as many factual mistakes as before.
- API access now costs half of GPT-5.6 rates, cutting bills for high-volume summarization and extraction work.
- Anthropic shipped Opus 5.5 minutes earlier, so expect faster model update cycles from both labs.
Why it matters
Anthropic shipped Opus 5.5 minutes earlier, so expect faster model update cycles from both labs.
This GenAI News article was prepared in original wording using reporting and materials published by TechCrunch AI. Source reference: https://techcrunch.com/2026/09/22/openai-launches-gpt-6-sol-and-luna/.
Drafted by the GenAI News review pipeline.