<h2>CONFIRMED: Claude Opus 5.5 reaches Claude Code with a 1 million token context</h2><p>Anthropic published Claude Code 2.1.280 on Tuesday, September 22, making Claude Opus 5.5 the new default model in the Opus family. This is more than a version bump. The announced price is $4 per million input tokens and $20 per million output tokens, with a 1 million token context window, putting the terminal among the launch product's main showcases.</p><p>What is confirmed is the <code>claude-opus-5-5</code> identifier in Claude Code. Anthropic also says Opus 5.5 is the default Opus model and that cache reads cost $0.20 per million tokens. That combination matters because Claude Code works with large repositories, session history, tools and files that quickly exceed the comfortable range of smaller contexts.</p><h3>What changes for Brazilian teams</h3><p>For a small or midsize company using agents for software maintenance, technical support or internal automation, a 1 million token window reduces the need to manually summarize long projects. It does not make heavy sessions free, but it makes it easier to keep documentation, code and previous decisions in one workflow. The advertised price also improves the economics of deeper reasoning, although the final cost in Brazil still depends on exchange rates, taxes, caching and the chosen provider.</p><p>The most interesting gain may come in repetitive pipelines. At $0.20 per million cached tokens, stable prompts, project instructions and repeatedly consulted files can cost much less than fully new calls. Teams still need to measure actual usage. A larger context is not automatically more efficient, and loading everything into every turn can increase latency and waste.</p><h3>The strategic signal</h3><p>Anthropic is not only selling a stronger model. It is trying to turn Claude Code into a continuous work environment where the model can follow an entire codebase and execute tasks with less context switching. Other changes in the release reinforce that direction, including MCP, plugins, background sessions, export and VS Code integration updates.</p><p>My read is simple: this launch matters because it combines capability, context and operations in one buying decision. The counterpoint is that independent benchmarks on real tasks are still missing, as is a public comparison of cost per completed task against competitors. For an SMB, the rational move is to test a representative workflow, measure tokens, time and rework, and only then change the default model.</p><h3>Sources</h3><p>Anthropic's official GitHub release: https://github.com/anthropics/claude-code/releases/tag/v2.1.280<br />Official repository changelog: https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md<br />Structured release record: https://api.github.com/repos/anthropics/claude-code/releases/tags/v2.1.280<br />Independent source consulted for contrast: https://vallettasoftware.com/blog/post/claude-opus-5-5-release-date</p>