On 1 July, Anthropic made Claude Sonnet 5 the default model for both free and paid users. That's not a footnote — Sonnet 5 performs close to Anthropic's flagship model on most tasks, and it's priced below the older Sonnet 4.6 through the end of August.
So if you've been paying for a premium AI plan specifically to get "the good model," the gap between what you get for free and what you pay for just got a lot smaller.
Why this is easy to miss
Most people don't actually check their AI subscription against what's free. You pick a plan once, based on whatever was best at the time, and never look again. Model quality moves fast. Subscription habits don't. Right now is a good moment to check whether you're still paying for something you don't need to.
Meanwhile, the paid side got more confusing
While Anthropic simplified things by giving everyone a strong free model, OpenAI went the other way. Its new GPT-5.6 launched split into three separate tiers — different names, different prices, different performance levels — with a rollout that isn't even fully public yet.
If you're building something on top of these tools, that's a real decision to weigh. If you're a casual user, it's mostly noise. The app will pick a tier for you.
What to actually do about it
If you're paying for AI mainly for quality: open the free tier this week and run your normal workload through it before you renew anything. If it holds up, that's money back in your pocket.
The takeaway: the biggest AI news this week isn't a new capability. It's a pricing shift that quietly moves value from paid plans to free ones — and that's worth two minutes to check.