With the Tuesday release of Claude Sonnet 4.6, Anthropic made its most powerful mid-tier AI model available to users with all subscription tiers, including free ones. Only a few weeks have passed since the company’s early February introduction of its flagship Opus 4.6 model.
The revised Sonnet model offers improvements in agent planning, long-context reasoning, computer utilization, and coding. The most significant change is the addition of a 1 million token context window in beta, which enables the model to handle dozens of research articles, lengthy contracts, or whole codebases in a single request.
Increased Cost And Access
The default model in Claude.ai and Claude Cowork for users on Free and Pro accounts is now Claude Sonnet 4.6. Starting at $3 per million input tokens and $15 per million output tokens, the pricing is the same as it was in Sonnet 4.5.
Additionally, Anthropic expanded its free tier to include previously paid-subscriber-only features like file creation, connectors, talents, and context compaction.
The model is accessible through the Claude API, Claude.ai, Claude Cowork, Claude Code, and popular cloud platforms including Google Cloud Vertex AI and Amazon Web Services Bedrock.
Reception Of Developers
According to Anthropic, early testing revealed that developers favored Sonnet 4.6 over its predecessor about 70% of the time. Long coding sessions were less frustrating for users since the model “more effectively read the context before modifying code and consolidated shared logic rather than duplicating it,” according to user feedback.
Notably, testers favored Sonnet 4.6 around 59% of the time over Opus 4.5, Anthropic’s frontier model released in November 2025. It was assessed as “significantly less prone to overengineering and ‘laziness,’ and meaningfully better at instruction following,” with more consistent follow-through on multi-step tasks and less hallucinations.
Advances In Computer Use
When compared to earlier Sonnet models, the model’s computer use capabilities have significantly improved. Anthropic said early users are seeing “human-level capability in tasks like navigating a complex spreadsheet or filling out a multi-step web form, before pulling it all together across multiple browser tabs”.
Since Anthropic first offered computer-use capabilities in October 2024, Sonnet models have steadily improved on OSWorld, the industry standard test for AI computer use.
Mike White, head of product at Anthropic, stated during the Opus 4.6 launch earlier this month that “Claude has changed from a model that could only help with minor tasks or answer questions to one that can handle substantial, significant work.”