OpenAI tuned GPT-5.6 Sol specifically for ChatGPT conversations, improving factual accuracy and response focus. Internal evaluation shows answers with at least one factual error were 68% less common with Sol compared to GPT-5.5 Instant. Plus and Pro users now have a slider to control reasoning effort per response, replacing the separate Instant and Thinking interfaces. For Free users, Luna is now the default model with unlimited text chats and a new Think button for questions requiring deeper reasoning.
On July 30, OpenAI cut prices significantly: GPT-5.6 Luna dropped 80% and GPT-5.6 Terra dropped 20%. Sol Fast mode (in the API) delivers up to 2.5x the speed of Standard processing at 2x the Standard price. Luna now costs 80% less but is performance-equivalent to models that were frontier-class a year ago; Terra remains competitive with GPT-5.5 at 40% of Sol's token cost. OpenAI credited Sol's own optimization work: Sol helped reduce end-to-end serving costs by 20% and improved speculative decoding efficiency by 15%.
These optimizations are being passed to all users—API, Codex, and ChatGPT. Auto-review in ChatGPT and Codex CLI upgraded from GPT-5.4 to Luna, expected to cost ~10x less. Both Luna and Terra support tools and multi-step workflows, making high-volume applications at scale more cost-effective. Model capabilities remain unchanged; the solver is just cheaper to run.
For architects: Luna's 80% cut changes the default routing logic for cost-conscious deployments. Evaluate whether tasks currently on Sol or Terra can drop to Luna with equivalent quality; A/B test the shift. The reasoning slider could unlock new UX patterns for user-facing chat (variable-latency responses based on question complexity). Watch whether other frontier labs match Luna’s pricing; if not, the cost-per-capability gap widens significantly.