OpenAI makes GPT-5.6 Luna default for all ChatGPT users, drops free-tier text limits

OpenAI announced this week that GPT-5.6 Luna becomes the default model across every ChatGPT tier, replacing GPT-5.5 Instant. The company's headline claim is a 62% reduction in responses containing at least one factual error, positioning Luna as a lightweight but more reliable everyday model. For the paid GPT-5.6 Sol experience, OpenAI cited a 68% reduction in factual errors on its high-stakes factuality evaluation spanning finance, medicine, and law.

The most consumer-visible change is the removal of daily message caps: free and Go tier users can now send unlimited text-only chats, though uploads of files and images remain rate-limited. Plus and Pro users received a new slider that lets them dial how long the model spends reasoning before answering — a user-facing exposure of the reasoning-effort knob that competitors like Anthropic and Google have been surfacing in their own products.

Competitively, the move lands squarely in a week defined by cost and access. DeepSeek is signaling price hikes, Alibaba and Meta are pressuring per-token economics, and OpenAI's answer is to make its cheapest reliable model free-with-no-limits — a defensive play to retain the massive free user base that feeds its funnel. AWS separately cut Bedrock's hosted GPT-5.6 Luna price 80% to $0.20 per million input tokens, underscoring how aggressively the Luna tier is being positioned on cost.

Watch whether 'unlimited' holds under load and whether the thinking-slider meaningfully changes output quality, or is mostly a perception feature. OpenAI also quietly retired GPT-4.5, consolidating those chats to GPT-5.5, continuing its steady model-version pruning.