The free tier got good

A current model as the default for people paying nothing, no chat limit — in the same week another lab raised a high-volume tier 93%. There is no price trend. There are four competitive positions.

GPT-5.6 Luna is now the ChatGPT free default with unlimited chats, following an 80% price cut. DeepSeek raised V4 Flash by 93% on 14 August.

Free tiers used to be where models went to be depreciated

Putting a current model there uncapped is a different decision, and it is only affordable because serving cost fell 80% two weeks earlier. The floor moved, not the ceiling.

At billion-user scale the differentiator is not the top of the range. A free tier that is genuinely good is harder to leave than a paid tier that is slightly better.

What that does to everyone else

Any product whose pitch was "better than what you get for free" has to re-establish it against a stronger free option. That is a large category of AI startups, and most of them found out this week.

93% up over here, unlimited and free over there. Neither number tells you what serving costs.

Where the war is actually being fought

Coding agents — enormous token volume, technical buyers who can evaluate in an afternoon, and integration shallow enough that switching is a config change. That is the definition of a winnable customer.

The unglamorous advice

Have an abstraction layer. The cheapest adequate model for a given workload is changing monthly, and price is behaving as a competitive variable rather than a technology curve — which means it can move against you.

Teams that can switch on a config line capture the discounting. Teams that wired a vendor into their architecture get to watch.

AI Weekly — AI News Today, August 15 — Top AI Stories & Live Updates → · LLM Stats — LLM News Today (August 2026) — AI Model Releases → · imFounder — AI Updates August 2026: 15 Explosive Stories to Know →