OpenAI launches GPT-6 Sol and Luna at half the price
OpenAI released GPT-6 Sol and Luna for ChatGPT, Codex and API users, targeting cheaper agent workloads with broadly similar performance.
What happened
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, making both models available in ChatGPT Work, Codex and the API. Sol is positioned as the balanced model for interactive and professional work, while Luna is the faster, lower-cost option for high-volume tasks. OpenAI says both models build on the alignment and reasoning advances introduced with GPT-6 Astra.
The commercial headline is pricing. OpenAI says Sol and Luna cost 50% less than comparable GPT-5.6 promotional pricing, while higher usage limits are available in several paid products. The company also introduced improved prompt caching for GPT-6, intended to reduce repeated input costs in long-running applications and agents.
Why it matters
The launch is less about a single benchmark breakthrough than about changing the economics of routine AI work. Cheaper inference makes it more practical to run agents across many small steps: reading files, calling tools, checking intermediate results and retrying failed actions. That matters more for software teams and enterprise automation than a modest improvement on a static leaderboard.
OpenAI's distribution strategy also gives the models immediate reach. GPT-6 Sol and Luna appeared in GitHub Copilot, Devin, Amazon Bedrock and Databricks shortly after launch, allowing developers to test them inside established workflows rather than through a standalone chatbot.
The unresolved question is whether the advertised savings survive real production workloads. Cache-hit rates, tool-call reliability and the cost of failed agent runs will determine whether Luna becomes a genuine default model or merely a cheaper option with more supervision overhead.