Collaborate with us Welcome back! Anthropic had a wild day. The US lifted the ban on Claude Fable 5, so its most powerful model comes back tomorrow, and it launched a new Sonnet 5 the same afternoon. The big labs also aimed AI at curing disease.
Google cut the price of AI video to pennies, and OpenAI halved what it costs to run ChatGPT. In today's Generative AI Newsletter: Claude: Why is its most powerful model suddenly back?
Claude Fable 5 beats the ban The most locked-down model in AI is coming back. Anthropic said Claude Fable 5 returns globally tomorrow, after the US Commerce Department lifted the export ban that pulled it offline last month. Fable 5 went dark on June 12, when Washington ordered every foreigner cut off from it, after Amazon found a way to jailbreak its cyber guardrails. Anthropic and the government spent three weeks building a fix.
New classifiers now catch the dangerous cyber tasks, and Anthropic says they block that jailbreak 99% of the time. There's a leash on it. To stay safe, Fable 5 will hand some routine coding and debugging to Opus 4.8 for now, and Anthropic will loosen the filters over the coming weeks. It's back on Claude.ai , Claude Code and Claude Cowork, for up to half your weekly usage through July 7, then on usage credits after.
And that was only Anthropic's second headline of the day. It also shipped Claude Sonnet 5 , its most agentic mid-tier model, near Opus quality at a fraction of the price with a 1M-token context window, already the default in Claude Code. The lab that spent a month locked out of its own model got it back and launched a new one before lunch. A few hours left before you lose the early investor perks Phase 1 of our raise closes in a few hours . The current share price and the best terms close with it.
Invest at an early investor price , starting at $1000 with up to 25% bonus shares. $2.8 million already in the funnel. GenAI Works owns distribution data nobody else can touch, the kind you get from helping Nvidia, Oracle and brands of every size reach millions. We created a product with this exclusive knowledge. ToneUp, an app that grows companies on data only we own.
Start small at $1000 Up to 25% bonus shares 2.5x YoY revenue in 2024 to 2025 14M member community Own a piece of GenAI Works before the window closes today. Invest today The AI labs came for science On the same day, OpenAI and Anthropic both aimed their models at science. Anthropic launched Claude Science, an app built for research. It traces every result back to the code that produced it, spins up computing environments on demand and connects to more than 60 scientific databases.
It's in beta. OpenAI answered with GeneBench-Pro , a test for whether AI agents can handle real computational biology, the judgment calls a working scientist makes instead of textbook questions. The point is drug discovery. An AI that can run an experiment could compress years of lab work into weeks and find medicines buried in data no human has time to read.
That is the biggest prize in AI, and it pays far better than chat. It's early days. Claude Science is a beta and GeneBench-Pro is a benchmark, so no one is curing anything this week. But both labs just showed you where they think the breakthroughs are, and it isn't your inbox.
Google made AI image and video cheap Google just dropped the price of AI media to pocket change. It launched two models for developers. Nano Banana 2 Lite makes an image in four seconds for about three cents, and Gemini Omni Flash generates up to ten seconds of video for ten cents a second, edited in plain English. The editing is the trick.
You can tell Omni Flash to swap a character, relight a scene or change the camera angle, and it keeps the original audio and motion intact. Google wants you to chain them, make an image with one and animate it with the other. At these prices, a small team can produce the kind of media that used to need a studio. OpenAI cut its AI bill in half OpenAI figured out how to run its models for less than half what it used to cost.
The Information reports OpenAI's engineers found new tricks that more than halved its inference bill. At one point it served ChatGPT to logged-out users on only a couple hundred Nvidia chips. This is the whole game. OpenAI ended the first quarter at a 39% gross margin and wants 52% by year-end, and cheaper serving is how it gets there, through higher usage limits, lower API prices or just fatter margins.
OpenAI is betting the next edge is how cheaply you can run a model, and it likes its odds against Anthropic there. Tool of the Day: Agno If you want to build an AI agent without gluing ten libraries together, Agno is the shortcut. It's an open-source Python framework for agents and multi-agent teams, model-agnostic, with memory, tools and a runtime called AgentOS built in. Try this yourself: Run pip install agno to install it, free and open source.
Build a starter agent in about five lines with a model, a tool like web search and a task. Add memory and a knowledge base so it remembers past runs and reads your docs. Build a team of agents that split a job and hand work off to each other. Who it's for: developers who want an agent running today, not a framework to wire up first.
Everything else you shouldn't miss Wall Street is leaning on AI to carry the quarter : Goldman Sachs says AI infrastructure will drive nearly 60% of the S&P 500's earnings growth this quarter, with Nvidia and Micron alone behind 40% of it. AI found a heart risk doctors can't see : UC Berkeley researchers trained a model to spot sudden-cardiac-death risk from a routine 10-second ECG, a warning no cardiologist had ever found in the waveform. Anthropic is giving Washington a first look at its models : Part of the deal to bring Fable 5 back has it handing the US government pre-release access to its models and safeguards, plus intel on jailbreaks. Nvidia says it cut DeepSeek's running cost to a fifth : New Blackwell software dropped DeepSeek V4's token cost about 80% in a month, with up to 20x more throughput on the same chip.
Learn more about AI from the experts building it 📸 Follow us on Instagram for fast, visual AI updates in 30 seconds. 🐝 Subscribe to our Atlas newsletter — trusted by 3M+ subscribers — to stay ahead of AI news across tech, education, and business. 📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators. 🐦 Follow us on X for breaking AI news and real-time industry updates. 🧠 Learn how to build your next AI application with practical resources and expert guidance. 🚀 Explore investment opportunities in the future of AI and join our community-backed growth journey.
Get in touch.