- Published on
Claude Code 2.1.284: Sonnet 5.5 Becomes the Default Model β 1M Context Is Now the Baseline
Thinking "it''s just a model name change" would be a mistake. Quadrupling the context window changes how you actually work.
Claude Code v2.1.284 shipped on September 28, 2026. The headline is simple: Claude Sonnet 5.5 is now the default Sonnet model. It looks like one number ticking up, but the practical difference is substantial.
Why the Default Model Change Matters
Most people using Claude Code never touch the model settings. The default is the everyday experience. That makes the choice of default model far more significant than it might appear.
Sonnet 5.5 stands out from the previous default in two key ways.
First, the context window is 1M tokens. A million tokens is hard to feel from the number alone. In practice, it means you can paste an entire mid-size codebase in one shot and say "fix this part." No more chopping files, resetting context, and re-explaining from scratch.
Second, it ships with Adaptive Thinking. Sonnet 5.5 supports "high effort" reasoning by default. However, Claude Code sets medium effort as the default β a deliberate balance between speed and quality. You can adjust the effort level in settings if needed.
Pricing and Performance: The New Math
Sonnet 5.5 API pricing is 10/MTok output. Prompt cache reads come in much cheaper at $0.20/MTok. For workflows that repeatedly use long context, caching can significantly cut costs.
The knowledge cutoff is June 2026 β recent enough to help with questions about the latest frameworks and libraries.
Other Changes in 2.1.284
Beyond the model swap, there are a handful of practical improvements.
- Terminal improvements: Better fullscreen mouse support for those who spend long sessions in the terminal.
- Stronger MCP controls: More granular management tools for MCP servers.
- VS Code session archiving: You can now archive sessions in VS Code to keep work history organized.
- Faster startup: Overall startup speed and response latency have improved.
How to Make the Most of This Update
Large file analysis is where Sonnet 5.5''s 1M context shines. Analyzing a legacy codebase in full, finding patterns across long log files, or reviewing a massive test suite all in one pass β these are the tasks that benefit most.
Non-coding work also sees the improvement. Writing long document drafts, large-scale data analysis, and complex multi-step reasoning all deliver more consistent results than before.
In team environments, it''s worth communicating the default model change to teammates. Teams using model-locking settings should verify how their existing configuration interacts with the new default.
Conclusion: The Baseline Just Got Higher
With Sonnet 5.5 as the default, users who never change any settings automatically get a 1M context window and adaptive reasoning. The "floor" of the tool has been raised.
In a world where AI coding assistants change every month, the quality of the default is the quality of the average user experience. This update raised that bar.
Sources: