Claude Sonnet 4.6 Rolls Out as Default AI Model — Near-Opus Power for Free Users

Anthropic has officially updated its AI lineup, making the highly capable Claude Sonnet 4.6 the new standard for everyone. This shift delivers significant performance gains to the free tier, offering logic and reasoning skills that nearly match those of the top-tier Opus model. Users can now receive faster, more sophisticated responses without a paid subscription.

On Tuesday, Anthropic launched Claude Sonnet 4.6, marking its second big AI release in less than a fortnight. The company claims this model is a step up for coding, design, and computer-based tasks, while also being better at handling heavy workloads and complex data.

Greater Capacity at No Extra Cost

A standout feature of this update is the new 1-million-token context window, now in beta, designed to help users manage large files and extensive reading material more easily. Claude Sonnet 4.6 has also become the standard model for anyone on the Free or Pro plans via claude.ai and Claude Cowork. Best of all, the price hasn't changed from Sonnet 4.5, staying at $3 (£2.21) for every million input tokens and $15 (£11.07) for every million output tokens.

According to a report by SSB Crack News, early adopters have already noticed a sharp improvement in the way Claude Sonnet 4.6 handles code. Developers are rapidly moving away from Sonnet 4.5 in favour of this new version because it more closely adheres to the specific instructions.

In fact, it often outperforms the high-end Claude Opus 4.5 from late 2025, demonstrating that Sonnet 4.6 can now handle the demanding office tasks that previously required a premium model.

Bridging the Gap in Computer Automation

This update is about more than programming; the model has become significantly better at using computers in general. This paves the way for automation in niche software that was previously too difficult to manage.

While earlier versions laid the groundwork for general computer use, Sonnet 4.6 has made a massive leap forward, according to the OSWorld benchmark. These tests, which mimic real-world software tasks, show the model navigating complex interfaces almost like a person—handling everything from tricky spreadsheets to lengthy, multi-step forms with ease.

As far as safety is concerned, rigorous testing confirms that Sonnet 4.6 is just as secure as previous models—if not more so. Researchers found that the model maintains a friendly, transparent, and helpful personality, backed by reliable safety protocols. Crucially, there were no signs of the major alignment problems that can sometimes affect high-level AI.

Setting New Benchmarks in User Preference

When tested, Sonnet 4.6 achieved impressive ratings across various industry benchmarks. This is supported by user testing in Claude Code, where 70% of participants selected Sonnet 4.6 as the superior model. They found it more adept at reviewing existing codebases and consolidating logic rather than duplicating it.

Feedback also suggests it is less likely to overcomplicate simple tasks, exhibits fewer hallucinations, and has a much better track record of following complex instructions.

With its large context window, the model can easily handle entire codebases or long documents, which is a significant advantage for projects that require careful, long-term planning.

Its knack for long-term strategy was proven during the Vending-Bench Arena evaluations. In those tests, Sonnet 4.6 prioritised building up its resources early on before pivoting toward profitability—a move that ultimately delivered much better results than its rivals.

Feedback suggests that the model's strengths are equally visible in creative design and financial tasks. Sonnet 4.6 consistently produces high-quality visual outputs, reducing the number of revisions required before a project is ready for use. In terms of document comprehension, its performance is now remarkably similar to that of Opus 4.6, making it an excellent tool for navigating and understanding large-scale corporate files.

A Smarter Ecosystem for Developers

Practically speaking, Sonnet 4.6's performance-to-price ratio is hard to beat, often serving as a great replacement for the more expensive Opus model. Its ability to generate clean iOS code and handle sophisticated logic makes it quickly become the go-to tool for developers working on extensive programming projects.

The Claude Developer Platform has also seen some clever updates, including new support for adaptive and extended thinking. By using context compaction, the system can now keep long conversations on track much more effectively. Additionally, new web search and filtering tools built into the API enable the model to handle search results more efficiently, resulting in higher-quality responses across the board.

Universal Availability

You can now find Claude Sonnet 4.6 across the entire Claude ecosystem, including the Cowork and Code apps, as well as on all the big cloud providers. The update also brings a significant boost to the free version of the service.

For those looking to integrate these features into their own projects, the model is available via the Claude API for immediate development and optimisation.