In partnership with
Hey,
Here is the sentence that made us read it twice: OpenAI pointed its flagship model, Sol, at OpenAI's own GPU code. Sol rewrote it. Serving costs fell 20%. Those savings were then handed to customers as a price cut of up to 80%.
The AI made itself cheaper to run. That is a new kind of headline, and it is the clearest signal yet of where this is going.
Three things changed what you pay this week: prices crashed, Claude's rate rises in 27 days, and California made AI labeling the law. Here they are.
1. The price crash
On July 30, OpenAI cut its cheap and mid tiers hard and left the flagship alone.
Model | Was | Now | Cut |
|---|---|---|---|
Luna (fast, cheap) | $1 / $6 | $0.20 / $1.20 | -80% |
Terra (workhorse) | $2.50 / $15 | $2 / $12 | -20% |
Sol (flagship) | $5 / $30 | $5 / $30 | unchanged |
Per million input / output tokens. Sol also gained a Fast mode: up to 2.5x quicker, at double the price.
What practitioners are reporting. Notion's AI product lead said Terra matched the previous generation's quality on scoped tasks at half the cost and in 60% less time. A founder building agents said Luna does the same work 40% faster and 40% cheaper than his previous default. The middle and budget tiers are now doing work that used to require the expensive one.
Why it happened. Chinese open-weight models have taken an estimated 46% of US enterprise token usage on one major routing platform, at a fraction of the price. Add enterprises finally auditing their AI bills, and price became the battleground. OpenAI also passed 1 billion active users the day after the cut, so this is a scale play, not distress.
Do this: if you build on the API, re-check which tier each task runs on. Plenty of work sitting on expensive models can drop a tier with no quality loss. If you pay for ChatGPT or Codex, you didn't get a discount, you got headroom: Terra and Luna now consume fewer credits, so the same plan buys more output.
2. The deadline nobody is flagging
27 DAYS LEFT
Claude Sonnet 5's introductory pricing runs through August 31. From September 1, input goes from $2 to $3 per million tokens and output from $10 to $15. That is a 50% increase, automatic, with nothing to opt into.
The wrinkle most coverage misses: Sonnet 5 ships a new tokenizer. Anthropic's own documentation says it produces approximately 30% more tokens for the same text, so an equivalent request can cost more even at unchanged per-token rates. Independent tests have measured higher on code.
If you build on Sonnet 5, this is the month to benchmark it against the newly cheap alternatives and decide on purpose. Consumer plans are unaffected.
3. AI labeling is now actually the law
The most-used AI image generator in the country is, as of this week, apparently out of compliance with a law that carries daily fines. Midjourney ships no C2PA content credentials and no known pixel watermark. That now matters.
California's AI Transparency Act became operative on August 2. Any generative AI provider with more than a million monthly users in the state must embed machine-readable provenance data in generated images, video, and audio, offer a free public tool letting anyone check whether content came from their system, and let users add visible AI labels.
Penalties are $5,000 per violation, and each day of continued noncompliance counts as a separate violation. A 30-day gap in your detection tool is $150,000 of exposure from that one deficiency. Enforcement sits with the Attorney General plus, unusually, city attorneys and county counsel, with fee-shifting.
Three things worth knowing:
This is effectively global. August 2 is the same day EU AI Act Article 50 transparency duties began applying. Nobody builds a California-only product.
Text-only output is not covered. Images, video, and audio are.
It is not fully settled. A pending urgency bill could still shift the details, so treat this as live rather than final.
The practical upshot for anyone publishing anything: "was this AI-generated?" is moving from an awkward question to a checkable fact.
Auto-Generate Free SEO Audit Report
Most e-commerce sellers have no idea why their listings aren't ranking. Wrong keywords, missing metadata, weak product descriptions — the problems are there, but no one's told you where to look.
StoreClaw runs a full SEO audit across your Amazon and Shopify stores automatically. In minutes, you get a clear score, a breakdown of what's hurting your rankings, and exactly what to fix.
No manual review. No SEO agency. No guesswork.
Connect your store and StoreClaw surfaces every issue that's costing you search visibility — then tells you how to fix it.
Free to start. No credit card required.
The ToolChase site has no ads, no pop-ups, and no cookie wall. Two labeled sponsors here are what pay for that, which also means our scores stay unbuyable and our reviews stay independent. If the partner above is useful to you, a click costs you nothing and keeps the whole thing running. Want one of the two slots? Here's how it works.
Also worth knowing
Claude's desktop app got a browser. Plus live shareable artifacts, so the interactive things Claude builds for you can now be handed to someone else as a working link. An iOS simulator beta and a Linux beta shipped alongside. The pattern across every assistant this year: they are growing hands, not just answers.
Meta AI went agentic. It can now plan, research, build presentations, and complete tasks through connected apps. Worth noting purely for reach: Meta AI sits inside apps billions of people already open daily, so for a lot of casual users this will be their first real agent experience, free.
The open tier keeps closing the gap. Alibaba released Qwen3.8-Max on August 3 at $2/$6 per million tokens, with 2.4 trillion parameters, a 1-million-token context window, and open weights promised within the week. DeepSeek's V4 Flash exited preview at $0.14/$0.28. Google's open Gemma family passed 900 million downloads. Every one of these pushes down what you will pay next quarter.
What we updated on ToolChase
All three of our most-visited tool pages moved this week. ChatGPT crossed a billion users and quietly got better value per subscription dollar. Claude gained the desktop browser. Perplexity remains the fastest way to check what another AI just told you.
The guides we refreshed off the back of this week's pricing news: Best Free AI Tools, Best AI Coding Assistants, and the flagship 50 Best AI Tools in 2026. Every tool page now carries the official product video, so you can watch a tool work before you spend a trial on it.
Own Search With Podcasts
Search engines and AI platforms reward brands that are mentioned, cited, and trusted across the web. PodPitch books your experts on relevant shows, creating branded mentions, backlinks, transcripts, citations, and reusable content. Only 20 demo spots are available this month. Once claimed, the offer disappears.
Prompt of the week: the bill audit
Prices dropped up to 80% and one major provider's rate rises in 27 days. Almost nobody adjusts. Paste this in and find out what you are leaving on the table:
You're a cost analyst for AI spending. Here's everything I currently pay for AI, tools, subscriptions, and API usage: [LIST: what it is, what tier, what you pay monthly].
1. Which of these am I on a higher tier than my usage justifies? Be specific about what I'd lose by dropping down.
2. Has anything here changed price, quietly raised rates, or introduced a cheaper tier in the last 60 days that I should switch to?
3. Am I paying two or more tools to do the same job? Which one should survive, and why?
4. Which of these should I absolutely keep even if expensive? Tell me plainly, don't force a downgrade that costs me time.
End with one number: my estimated annual saving if I act on your recommendations.
Why it works: AI pricing now moves faster than anyone's renewal cycle. The deal you signed six months ago is almost never the deal available today, usually better, occasionally worse. Fifteen minutes here, once a quarter, is the highest hourly rate you will earn all month.
That is the signal for this week. Capable AI keeps getting cheaper faster than anyone's habits update. Go check your tiers.
Reply and tell me one thing: what's the most you pay for a single AI tool, and is it worth it? I read every reply, and the honest answers are going into a future issue.
See you next Wednesday,
Emre
Founder, ToolChase
Building an AI tool? Submit it for a free editorial review. Read The State of AI Tools 2026, our study of the full directory.