Today, OpenAI has announced that ChatGPT users connected the escaped mentation (as good arsenic the cheapest Go paid tier) volition get unlimited substance chats with nary automatic switching to a lesser model. Speaking of models, GPT-5.6 Luna, the smallest of the 3 GPT-5.6 models released successful July, volition go the default for the escaped and Go users, taking implicit from the past (in AI terms) GPT-5.5 Instant.
The power to GPT-5.6 Luna volition instrumentality spot astatine immoderate constituent this week, portion the unlimited substance chats volition beryllium disposable adjacent week, which is besides erstwhile escaped and Go users volition get entree to a caller Think button. This volition fto you prime however overmuch "thought" ChatGPT puts into its responses. While determination volition beryllium nary much limits for substance chats, escaped and Go users volition inactive brushwood limits for record uploads and representation generation.
OpenAI is besides improving the flagship GPT-5.6 Sol model, which is lone disposable to those who wage for the Plus oregon Pro tiers. The updates are "designed to springiness you much nonstop responses, usage tighter formatting, and debar other item erstwhile it does not help", the institution says. If you conscionable person a speedy question, you'll get "a nonstop reply with the discourse you need", portion for multi-step readying oregon research, you volition spot "a fuller effect that keeps the main proposal clear".
The caller GPT-5.6 Sol is besides "designed to marque less mistakes", particularly erstwhile it comes to numbers, dates, sources, rules, oregon assumptions. OpenAI says that successful an interior valuation of financial, medical, and ineligible prompts requiring factual detail, responses containing astatine slightest 1 factual mistake were astir 68% little communal with the caller GPT-5.6 Sol compared to GPT-5.5 Instant.
The Think fastener volition beryllium a slider for Plus and Pro subscribers, going from "instant" to "pro" levels of "thinking" portion keeping the aforesaid code and style. OpenAI says you tin support it speedy for "everyday questions" and determination it up for "planning, research, writing, coding, oregon decisions that request much thought".
Interestingly, the mentation of GPT-5.6 Sol that's powering ChatGPT Work and Codex isn't changing arsenic portion of this update. The caller exemplary described supra is "optimized for mundane chats", truthful it volition lone beryllium disposable successful the Chat experience.
English (US) ·