GPT-5.6 Sol Cuts Factual Errors by Up to 68%: Here’s What Changed in ChatGPT.
OpenAI Is Going Unlimited on Chat and Physical With a $300 Donut:
Free unlimited text chats, a smarter GPT-5.6 lineup, and a screenless AI speaker signal the same strategy: get AI in front of more people, in more places, more of the time.
1B+ Weekly ChatGPT users
68%: Fewer factual errors in GPT-5.6 Sol
$300–400: Rumored price of OpenAI's speaker
OpenAI had two announcements land in the same news cycle, and together they tell one story: after crossing a billion weekly users, the company is racing to make ChatGPT harder to hit a wall with and harder to walk away from.
First, OpenAI is removing text chat limits for every ChatGPT user, free or paid. Second, fresh reporting fills in more detail on the company's first hardware device — a donut-shaped, screenless AI speaker built with Jony Ive's design studio LoveFrom. One move expands access to the software; the other extends that access into physical space.
1: No more counting your questions:
Free and Go users are moving to a new default model, GPT-5.6 Luna, replacing GPT-5.5, and next week they'll get unlimited text-based chats along with a new "Think" button for tougher questions that need more reasoning. Separate limits still apply to files, images, voice, and image generation, so the unlimited promise is specifically about text conversation.
Plus and Pro subscribers aren't left out. They're getting an upgraded GPT-5.6 Sol model tuned for quick tasks — research, advice, planning, writing, decisions — along with a thinking slider that lets them dial reasoning effort up or down based on how hard the question actually is. OpenAI says internal testing found factual errors dropped 62% for GPT-5.6 Luna and 68% for GPT-5.6 Sol compared to GPT-5.5-Instant, a meaningful accuracy jump for models now serving well over a billion weekly conversations.
The Hidden AI War
Nobody Is Telling You About
Our latest documentary deep-dive into the geopolitical struggle for machine intelligence dominance. Explore the two paths of AI development: open source vs. closed architecture.
Factual errors are down 62% for GPT-5.6 Luna and 68% for GPT-5.6 Sol versus GPT-5.5-Instant, per OpenAI's internal evaluation.
2: An AI speaker with moving parts:
On the hardware side, Bloomberg reports OpenAI's long-rumored device is a donut-shaped, screenless speaker designed to be carried from room to room — bedside table one moment, kitchen counter the next. It's said to be built from "high-quality metal" with a premium finish and, intriguingly, actual moving parts. Pricing is expected in the $300 to $400 range, well above most of Amazon's smart speaker lineup, which tops out closer to $240.
The strategic logic is straightforward: a physical, always-present device would weave ChatGPT further into daily routines than an app ever could. The risk is just as straightforward — smart speakers have a rocky profitability history, a premium price tag raises the bar for adoption, and OpenAI is developing the device while facing a trade-secret lawsuit from Apple. A 2027 release is expected.
3: The pattern behind both moves:
Removing usage limits and building standalone hardware look like different bets, but they're solving the same problem: friction. Every cap on a free conversation, every extra tap to open an app, is a moment where a person might not come back. OpenAI is betting that the fewer barriers between a person and the model, the more that model becomes infrastructure rather than a tool people remember to open.
That's the same bet worth making inside a business. AI that's rationed, gated behind a dashboard, or bolted onto one department doesn't become part of how the company runs — it stays a tool a few people remember to use. AI that's simply always available, in the channel someone already works in, is what actually changes day-to-day operations. Remove the friction from AI at your business, not just your chatbot.
OpenAI is stripping away limits and adding new surfaces because unrestricted access is what makes AI stick. Otherworlds AI's Agent+ platform applies that same principle to your operations — always-on, no per-seat rationing, handling leads, requests, and workflows automatically for $297/month. Need it built around a specific process? Our custom enterprise AI builds start there.
Learn more at otherworldsai.com
