Anthropic released Claude Opus 5 on 24 July 2026, and unlike most model launches, it didn’t sit behind a manual opt-in for long. Claude Code version 2.1.219, shipped the same day, made Opus 5 the resolved default for Max subscribers, Team Premium, Enterprise pay-as-you-go, and API users calling the opus alias. It’s a live model swap inside a tool teams already have running in CI pipelines, IDE integrations and terminal workflows — not a new product to evaluate, but an existing one that started behaving differently overnight.
The rollout has a catch worth knowing before you go looking for it in your own account. On Pro and Team Standard plans, the default alias still resolves to Sonnet 5, not Opus 5 — Anthropic drew the line at pricing tier rather than rolling it out universally. That means two developers on the same team, one on Max and one on Pro, can run identical /model default commands in Claude Code today and get two different models back. Early users have also reported behavioural differences beyond raw capability: Opus 5 verifies its own work without being asked more often than Opus 4.8 did, tends to write longer responses by default, and reaches for subagents more readily on multi-step tasks — all of which change token usage and session cost, not just output quality.
So what
If you’re commissioning or managing an AI-assisted development engagement, this is a reminder that “we use Claude Code” describes a moving target, not a fixed tool — the model underneath, the default behaviours, and even which plan tier gets which model are all live variables that shift on Anthropic’s schedule, not yours. It’s also a genuinely practical thing to check this week: teams on Pro or Team Standard plans who assumed they’d automatically get Opus 5’s improvements haven’t, and teams on Max who have will see cost and verbosity shift whether they asked for it or not. Managing that drift — knowing which model is actually running a given task, and whether that’s the right call for the work — is exactly the kind of governance question that separates a considered AI-assisted development setup from one that’s quietly drifting. If you want a second opinion on how your team’s AI coding tooling is actually configured, get in touch.