A Busy Week at Anthropic

Anthropic, the company that makes the AI assistant Claude, had a busy week. On April 22, it made a new version of its top model, Claude Opus 4.7, generally available. That means anyone with an account, and not just early testers, can now use it. The new model is better at writing computer code and at completing long tasks without close human help. It can also read images at a higher quality than before. The price, however, is the same as the older version: five dollars per million words sent in, and twenty-five dollars per million words written out.

On the same day, Anthropic ran a small experiment with its pricing. For about two percent of new users on the cheaper Pro plan, the company quietly removed Claude Code, a tool that lets people write software directly with the model. The aim was to see whether new buyers would still sign up without that feature. The reaction online was strong, especially from developers who depend on the tool, and Anthropic quickly reversed course. By the end of the day, Claude Code was back in the Pro plan for everyone. The episode showed how closely developers now watch every change to AI pricing.

The company also reminded the public that Opus 4.7 is not its most capable model. A more powerful system, called Mythos Preview, has only been shared with a handpicked group of security companies. Anthropic has held it back because it is unusually strong at finding weaknesses in computer systems, and the company wants to use it first to help defenders rather than risk it helping attackers. This careful, slow release is becoming a pattern at Anthropic, which often argues that safety should set the speed of new technology.

Finally, on April 27, Anthropic opened an office in Sydney and named a head for Australia and New Zealand. The move shows how quickly the company is growing outside the United States, where most of its staff still work. Taken together, the week was a small picture of where the AI industry is right now: faster and more capable models, careful pricing experiments, public debates about safety, and new offices opening in big cities almost every month. None of this looks likely to slow down in the months ahead.