Voltar

Anthropic launches Sonnet 4.6 with expanded context window and benchmark gains

Model Release Overview

Anthropic announced the release of Sonnet 4.6, continuing its regular four‑month cadence of model updates. The new version is positioned as the default offering for users on both the Free and Pro subscription tiers, signaling a broad availability across the platform’s user base.

Technical Enhancements

Sonnet 4.6 brings notable improvements in three key areas: coding assistance, instruction following, and computer use. These enhancements aim to make the model more effective for developers, knowledge workers, and users requiring advanced task execution.

In the beta phase, Sonnet 4.6 introduces a context window capable of processing one million tokens. This capacity is described as twice the size of the largest window previously offered for Sonnet models, providing enough space to hold entire codebases, lengthy contracts, or dozens of research papers within a single request.

Position Within Anthropic’s Portfolio

The release arrives just two weeks after Anthropic launched Opus 4.6, its larger‑scale model. An updated version of the Haiku model is expected to follow in the coming weeks, indicating a coordinated rollout of the company’s model family.

Benchmark Performance

Sonnet 4.6 achieves a new set of benchmark scores. It records top performance on the OS World benchmark for computer use and on SWE‑Bench for software engineering tasks. Most prominently, the model attains a 60.4% score on the ARC‑AGI‑2 benchmark, a metric designed to assess capabilities linked to human‑level intelligence. This score places Sonnet 4.6 above most comparable models, though it still trails Anthropic’s Opus 4.6, Google’s Gemini 3 Deep Think, and a refined version of OpenAI’s GPT 5.2.

Implications for Users

By making Sonnet 4.6 the default model for both free and paid tiers, Anthropic ensures that a large segment of its user community can immediately benefit from the expanded context window and the refined performance in coding and instruction tasks. The beta availability of the extended token window suggests that developers and researchers will be able to work with substantially larger inputs without needing to split content across multiple requests.

Future Outlook

Anthropic’s rapid succession of model releases, highlighted by the close timing of Sonnet 4.6 and Opus 4.6, points to a strategic emphasis on continuous improvement and diversification of its AI offerings. The anticipated Haiku update will further round out the portfolio, providing options tailored to different performance and cost requirements.

Usado: News Factory APP - descoberta e automação de notícias - ChatGPT para Empresas

Source: TechCrunch

Também disponível em: