OpenAI’s Tibo Sottiaux posted on X that GPT-5.6 Sol is getting a set of cost and speed changes, including “80% lower prices on Luna,” “20% lower prices on Terra,” faster speeds for GPT-5.6 Sol in “/fast mode,” and a roughly “10X cheaper” auto-approve mode — labeled “review for me” in the app — that now uses Luna to “prevent many high risk actions” from the main agent.
Sottiaux later added that the changes affect both API pricing and usage in paid subscriptions. The quoted OpenAI post attached to the thread describes the rollout as a broader effort to push “cost efficiency, capability, and speed,” while also noting that Luna and Terra’s lower prices will be reflected in how usage is counted in Codex and ChatGPT Work.
One accompanying graphic, titled “Artificial Analysis Intelligence Index v4.1,” leans into the pricing pitch by placing GPT-5.6 Luna at the top of a cost-versus-intelligence chart. Its subtitle claims Luna “achieves the highest intelligence score at a fraction of the cost of similarly capable models.” The chart compares Luna against models including DeepSeek V4 Pro, MiniMax Default, Gemini 3.5 Flash-Lite, GLM-5.2 Max, Gemini 3.1 Pro Preview, Claude Opus 5 Low, Gemini 3.6 Flash, Claude Sonnet 5 High and Claude Haiku Reasoning. As with any vendor-published performance graphic, that comparison should be treated cautiously.
Source: Tibo on X
