OpenAI puts its cheapest model on the front door: Luna becomes the free ChatGPT default, with the meter switched off

OpenAI announced on 6 August 2026 that GPT-5.6 Luna becomes the default model for ChatGPT's Free and Go tiers, replacing GPT-5.5, and that text conversations on those tiers are no longer rate-limited. A Think button, which spends extra reasoning effort on a single message, follows from the week of 10 August. Paid Plus and Pro users move to the updated GPT-5.6 Sol with a reasoning slider running from Instant to Extra High. The interesting part is the arithmetic. Luna is the entry tier of the GPT-5.6 family: it carries the same 1.05-million-token context window and the same 128,000-token output ceiling as Sol, but OpenAI's own price list puts it at $0.20 per million input tokens against Sol's $5. Twenty-five times cheaper input is what makes an unmetered free tier a business decision rather than a giveaway - the company is not absorbing frontier costs, it is routing free users to the model that costs it least. The word "unlimited" also does less work than it looks. The cap comes off text chats only; file uploads, image generation and other tools keep their existing limits, and OpenAI notes the offer remains subject to abuse guardrails. On quality, OpenAI's internal testing claims Luna makes 62 per cent fewer factual errors than the model it replaces on legal, medical and financial prompts. That is a vendor number, measured by the vendor, on the vendor's prompts - worth recording, not worth treating as settled.