Anthropic Expands AI Competition With Claude Sonnet 4.6 and a 1 Million Token Context Window

Anthropic has intensified the race in advanced artificial intelligence models with the release of Claude Sonnet 4.6, a system the company describes as its most capable Sonnet-tier model to date. The update introduces a 1 million token context window in beta and narrows the performance gap between the mid-tier Sonnet line and the company’s premium Opus models.

With this expansion, users can process entire codebases, extended legal contracts or large research archives within a single prompt. The move directly addresses one of the most pressing demands in enterprise AI adoption: sustained reasoning across massive volumes of information without fragmenting tasks into smaller queries.

According tospan>Anthropic/span>, Claude Sonnet 4.6 brings substantial upgrades in coding, long-context reasoning, agentic task planning and computer use. The company also claims improved design fluency and enhanced resistance to prompt injection attacks, a growing security concern as language models are increasingly embedded into enterprise workflows.

Notably, pricing remains unchanged from Sonnet 4.5, set at $3 per million input tokens and $15 per million output tokens. Sonnet 4.6 is now the default model for both Free and Pro tiers, signaling Anthropic’s intention to democratize higher-tier performance rather than confining it to premium plans.

The company says capabilities that previously required an Opus-class model are now accessible within the Sonnet tier. Early user testing appears to support that claim. In internal comparisons, users reportedly preferred Sonnet 4.6 over Sonnet 4.5 roughly 70% of the time and even selected it over Claude Opus 4.5 in 59% of evaluations. Feedback cited stronger instruction adherence, fewer hallucinations and less unnecessary complexity in responses.

Performance gains extend beyond text generation. On the OSWorld benchmark, which evaluates real-world software interaction without APIs, Sonnet 4.6 approaches human-level performance in tasks such as navigating spreadsheets and executing multi-step web workflows. These capabilities reflect a broader industry shift toward AI agents capable of interacting directly with digital environments rather than merely generating text.

The release comes just weeks after Anthropic unveiled Claude Opus 4.6, its most advanced model to date. By elevating Sonnet to near-Opus performance while maintaining lower pricing, Anthropic appears to be recalibrating its product segmentation strategy. The goal is clear: compress the distance between cost and capability while expanding practical enterprise use cases.

As competition intensifies among major AI labs, context window size and agentic reliability are emerging as key battlegrounds. A 1 million token capacity positions Sonnet 4.6 among the most expansive commercially available models, particularly relevant for developers and enterprises working with dense documentation or complex code repositories.

In a market increasingly defined by incremental improvements, Anthropic’s latest move signals a strategic pivot toward scalability and accessibility. By upgrading its mid-tier model to deliver performance once reserved for flagship systems, the company is betting that broader adoption, rather than exclusivity, will define the next phase of AI competition.