Claude Code just made Opus 5 the default — 1M context window, and fast-mode pricing that changes the calculus
On 24 July 2026, Anthropic shipped Claude Code 2.1.219, making Claude Opus 5 the default Opus model with a 1M-token context window and $10/$50 per-Mtok fast-mode pricing — a meaningful jump for anyone tracking 'Claude Opus 5' and 'AI coding tools 2026' search terms as a proxy for which AI vendor to standardise on.
29 July 2026
Anthropic’s Claude Code changelog for 24 July 2026 (version 2.1.219) confirms Claude Opus 5 is now the default Opus model in Claude Code, replacing Opus 4.8. The headline spec is a 1M-token context window — roughly four times what most teams have been working with day to day — alongside fast-mode pricing of $10 per million input tokens and $50 per million output tokens. Anthropic also quietly retired Opus 4.7 from /fast mode entirely, so fast mode now only applies to Opus 5 and Opus 4.8. The same release raised the default nested-subagent spawn depth from 1 to 3, letting orchestrated agent workflows delegate sub-tasks further before a human needs to step back in.
None of this is an incremental point release. A 1M-token context window means an agent can hold a genuinely large codebase, or a full requirements document plus the relevant service layer, in working memory at once — the kind of context loss that has historically forced developers to re-explain constraints mid-task simply stops happening as often. Combined with deeper subagent nesting, it’s a meaningful step toward AI coding tools that can be handed a larger, less tightly-scoped piece of work and trusted to keep the thread.
So what
If you’re a founder or CTO benchmarking “which AI coding tool should our team standardise on” — a search that spikes every time a frontier lab ships a new default model — this is the kind of release that should move the needle on that decision, not just the marketing headline. A bigger context window and deeper agent delegation reduce the babysitting overhead that’s been the real cost of adopting these tools, which is exactly the kind of operational detail that separates AI-assisted development done well from AI-assisted development done as a gimmick. If you’re weighing how to build model-agnostic tooling or governance around which agents your team runs, that’s the work we do in AI-assisted development engagements — get in touch if you want a second opinion before you commit budget to a particular stack.