A quiet weekend, so this is a short edition. Two things happened that actually change what you can do, and nothing has been padded out to fill a list.
Free ChatGPT now defaults to GPT-5.6 Luna
Free ChatGPT moves to GPT-5.6 Luna with unlimited text chats, and Claude Fable 5 stops handing off everyday biology questions.
GPT-5.6 Luna becomes the free default
OpenAI is making Luna the default model for Free and Go accounts.
Luna is the fastest and cheapest model in the GPT-5.6 family.
A Think button follows the week of August 10.
- Unlimited text chats on the free tier
- Factual errors roughly 62% less common than GPT-5.5 Instant, on OpenAI's own evaluation
- File uploads, images and other tools stay capped
- Announced August 6, rolling out this week
The free tier is where most of your audience actually lives, and it just got materially better.
Any "which AI should I start with" content you have published now gives the wrong answer, because the honest one stopped being "pay first".
Improving GPT-5.6 Sol in ChatGPTClaude stops deflecting biology questions
Anthropic rewrote the constitution of the classifier that decides what counts as safeguarded biology.
It was retrained with internal and external expert feedback.
A fallback is the automatic handoff that routes a query to a less capable model.
- Biology-related fallbacks down roughly 85% in testing
- Virology, toxicology and molecular design still fall back to Opus 5
- Published August 7
If you make health, fitness or science content, the refusals that used to derail a research session on lab results or symptoms largely stop.
Anything close to dual-use work is still handed off, so this is not a green light for research-grade biology.
Improving Fable 5 SafeguardsWe are making updates to Claude Fable 5 biology safeguards in a way that substantially reduces fallbacks.FAQs about this AI news edition
Here are answers to the most common questions about these two releases.
GPT-5.6 Luna. OpenAI announced on August 6 that it becomes the default for Free and Go users, replacing GPT-5.5 Instant.
The week of August 10, alongside a new Think button. Limits stay in place for file uploads, images and other tools.
On OpenAI own evaluation of financial, medical and legal prompts, replies containing at least one factual error were about 62% less common on Luna than on GPT-5.5 Instant.
Everyday ones, yes. Anthropic cut biology related fallbacks by roughly 85% in testing on August 7. Virology, toxicology and molecular design still fall back to Opus 5.
The automatic handoff that routes your query to a less capable model when the system judges it touches safeguarded territory.
Navid Moazzez
Founder of Creator School AI & AI Creator OS
Navid Moazzez helps creators and founders master AI and build their own AI Operating System (AI OS) to automate their business and life.
🤖 Free virtual event
Want insider access to 50+ exclusive AI creator workshop sessions?
Thrive in the creator economy in the new age of AI and learn from leading-edge AI experts, creators, and entrepreneurs who've been in the trenches for years.
Claim your free access to this actionable content today.
P.S. Register to instantly get my ultimate AI tools guide for creators!
Loved by 100,000+ previous attendees