Anthropic and OpenAI Launched New Models on the Same Day. Then Both Handed Out Free Usage.

1  Anthropic's Opus 5.5 matches Fable-level performance at 40% less cost, and it finally sounds less robotic.

The newest Opus 5.5 delivers roughly the same capability as Anthropic's top-tier Fable models while costing 40% less than Opus 5. The more noticeable change for everyday users may be the writing itself: the model produces more natural prose, a direct response to one of the most repeated complaints about recent Claude releases. It's live now across every Claude tool, and Anthropic published a dedicated prompting guide to help you get more out of it from day one.

2  OpenAI split its GPT-6 lineup into cheaper tiers, one built for heavy reasoning and one built for speed.

GPT-6 Sol handles demanding work that needs serious reasoning, while GPT-6 Luna is tuned for everyday tasks where speed and cost matter more than depth. Both are rolling out now across most ChatGPT Work and Codex plans, and API prices dropped by half alongside them. Having a clear cheap tier and a clear heavy tier makes it much easier to match the model to the job instead of paying flagship prices for routine work.

3  Mirage built a video editing suite designed for AI agents rather than people.

Tesseract gives agents access to the same building blocks a human editor uses: compositions, keyframes, and audio tracks. You can watch a preview while the agent works, ask for changes along the way, and render the final cut locally on your own machine. It works inside both ChatGPT and Claude today. See a walkthrough here.

Same day, same playbook. Anthropic and OpenAI shipped major model updates within hours of each other, and both paired the launch with something extra for paying users. Anthropic raised its five-hour usage limits on paid plans and added a rate limit reset you can trigger whenever you want. OpenAI added a banked usage limit reset of its own and cut API pricing in half. Neither company needed to do this to prove its model was good. They did it to keep people from switching.

Price is now a feature. A year ago, the pitch was simply that each new model was smarter. This week, both headlines led with cost: 40% cheaper for Opus 5.5, 50% cheaper API calls for GPT-6. When the two strongest labs lead with savings instead of benchmarks, it's a sign that capability gaps at the top have narrowed enough that price and usage limits are becoming the real tiebreakers.

The stakes behind the discounts. There's a bigger reason both labs are fighting this hard for loyal users right now. The Financial Times estimates that public listings for SpaceX, OpenAI, and Anthropic could together be worth more than every tech IPO since 1980. Heading into that kind of moment, active users and retention numbers matter enormously, and handing out extra usage is a cheap way to protect both.

What it means for you. For now, this competition works in your favor. More usage, lower prices, and better models are arriving faster than most teams can adopt them. The practical move is to avoid locking into a single lab: test both new lineups on your real tasks this week, and let the one that does the job best at the lowest cost earn the work.

What's trending on socials and in the headlines today

📝 The official playbook. Anthropic's own guide to getting the most out of Opus 5.5 collected 5,000 bookmarks in its first two hours. It walks through prompting the new model, keeping long tasks on track, and checking the results before you trust them.

🔮 A skeptic converted. One self-described Meta critic admitted Muse won him over after the agent secured a $250 credit on his behalf following a seven-hour flight delay. He called the experience "magic," which is not a word Meta hears often from its detractors.

🤑 A fund with no traders. Bracket22 is a hedge fund where AI agents make every investment decision. Its manager, Brian Kelly, told CNBC his main job is occasionally pressing a button. The interview clip is a sharp illustration of how quickly entire professions are being rewired.

📖 A model built for storytellers. A new lab released the Narrative World Model, designed to plan and write complex story arcs from the first chapter to the last. The launch post passed 6 million views. Try writing with it here.