Cheaper Agents Just Landed

Anthropic launched Claude Sonnet 5, and the headline is price. It is built to run AI agents, meaning software that does multi-step tasks on its own, at a much lower cost than top-tier models.

An agent is basically an AI that can take action, not just chat. Think booking, sorting, researching, or moving data between apps without you babysitting each step. Until now, running those agents on the strongest models got expensive fast.

The numbers tell the story. Sonnet 5 runs at $2 per million input tokens and $10 per million output as an introductory rate through August 31, then settles at $3 and $15. For that price Anthropic says it performs close to Opus 4.8, its top-tier model, while beating the older Sonnet 4.6 on reasoning, tool use, and coding. It is now the default model on the free and Pro plans.

There is a safety angle too. Anthropic reports Sonnet 5 shows fewer undesirable behaviors than Sonnet 4.6 and is generally safer to let run on its own, which matters the moment you hand an agent real tasks.

A startup called Acti is putting AI agents directly into your smartphone keyboard on iOS and Android. The idea is simple: the keyboard is the one place you already type in every app, so why not let AI help there too.

The keyboard works across apps and lets you build custom shortcuts using plain language. You describe what you want, and it sets up a quick action you can trigger while texting, emailing, or posting.

This is part of a bigger shift. Instead of opening a separate AI app, the assistant comes to wherever you are already working.

Google updated its image generator with Nano Banana 2 Lite, a version made to be faster and cheaper. The goal is to make AI image creation more practical for everyday creators.

If you make marketing graphics, listing photos, or social content, speed and cost are what stand between testing an idea and shipping it. A lighter, cheaper model lets you generate more variations without watching the meter.

The trade is usually some quality for speed, but for quick drafts and volume work, that can be exactly the right deal.

OpenAI released new Signals data showing how ChatGPT adoption is growing across the globe. People are not just signing up; they are using it more often and trying more of what it can do.

The growth spans more regions and languages, which means your customers, employees, and competitors are increasingly comfortable with these tools by default.

For a builder, the takeaway is less about OpenAI and more about the baseline. AI fluency is becoming normal, so leaning on it in your own work no longer feels early or strange to the people you serve.

OpenAI published a report mapping how AI might change work across the European Union. It highlights which jobs could face automation, which could grow, and which will mostly see their daily workflow shift.

The useful framing here is that not every job gets replaced. Many simply change shape, with AI handling routine parts and people focusing on judgment, relationships, and decisions.

Even if you are not in the EU, the categories are a handy lens. Look at your own team and ask which tasks are routine enough to hand off and which are the human core worth protecting.