Claude Code Closed to Minors, Qwen Code Goes Paid, and Opus 4.7 Uses 30% More Tokens

It was a busy week for AI tooling: Anthropic introduced age restrictions for Claude Code, Alibaba moved Qwen Code to a paid model, OpenAI expanded Codex's capabilities, and the release of Opus 4.7 raised questions about the rising cost of using top-tier models.

Claude Code closed to minors

Anthropic updated the Claude Code user agreement, and the service now requires age confirmation. Users under 18 cannot use the tool without additional permission from a parent or guardian.

The move makes sense given growing regulatory pressure: COPPA in the US and similar initiatives in the EU are forcing AI companies to build protections for minors. The phrase "requires a passport" is colloquial, but the substance is accurate: age verification is now mandatory.

Qwen Code: the end of the free era

Alibaba has ended the free period for Qwen Code, an AI coding assistant built on the Qwen model. The service is moving to a paid model. Exact pricing has not been disclosed yet, but it is a signal: major providers are abandoning the "attract first, monetize later" approach.

Qwen Code competed with GitHub Copilot and Cursor precisely because it offered free access to a strong model. That argument is now gone, and users will have to weigh which option is the better value.

Codex: works with any application

OpenAI expanded Codex, its AI agent for development. The key update: the agent can now control a computer directly, interacting with practically any application through a GUI. Previously, Codex worked mainly in the terminal and with the file system. Now it can open a browser, click through interfaces, and work with desktop programs. That is a step toward a full computer-use agent, not just code completion.

The update puts Codex alongside Anthropic Computer Use and Microsoft Copilot in agentic control mode.

Opus 4.7: more capability, more tokens

Anthropic released Opus 4.7, an update to its flagship model. Users noticed one key characteristic right away: the new version consumes 30% more input tokens than Opus 4.6. That means higher costs for anyone using the model through the API. In return, it offers improved reasoning quality and expanded context handling.

For Claude Code Max subscribers the difference is less noticeable, but for API customers with heavy workloads, the update calls for a budget review.

What this means

The four stories from one week add up to a single trend: the free era of AI tools is ending, and the leading players are tightening access rules and raising price levels at the same time. Age restrictions are a regulatory inevitability. Qwen Code's monetization and Opus 4.7's higher token consumption are business decisions. Codex's expansion is a race for the status of a full computer-use agent.

For development teams, this is a moment to rethink the tooling stack: what to use, what to subscribe to, and what to build into the budget.