Claude Sonnet 4.5 Is Released

Anthropic unveiled its latest AI model, Claude Sonnet 4.5, on September 29, 2025, positioning it as the leading model for coding and intensifying the competitive landscape with rivals like OpenAI and Google.

More than a simple upgrade, this release introduces new architectural designs aimed at enabling AI systems to operate as independent, long-term collaborators. A key innovation is the model's capacity for over 30 hours of continuous, autonomous work on intricate tasks, a substantial increase from the seven-hour limit of its forerunner, Claude Opus 4. This extended operational ability allows Sonnet 4.5 to handle long-term projects like full software development lifecycles and in-depth research with minimal human intervention. This leap in autonomy is supported by several new features: a beta Memory Tool that offers virtually limitless context by saving information across sessions, an automated context-clearing function that can cut token use by up to 84%, and internal tracking of token limits to avoid task interruption. These advancements transform the AI from a conversational tool into an independent digital workforce, reshaping the potential of enterprise automation.

Anthropic supports its assertions with top-tier results on industry benchmarks that measure practical application. The model scored 77.2% on SWE-bench Verified, a benchmark for real-world software engineering challenges, outperforming competitors such as GPT-5 Codex (74.5%) and Gemini 2.5 Pro (67.2%). Furthermore, it achieved a new high score of 61.4% on OSWorld, a test of an AI's capability to execute tasks on a computer, highlighting its enhanced agentic functions. The emphasis on such tangible, value-driven benchmarks indicates an industry-wide move toward evaluating AI on its ability to perform concrete tasks.

Qualitative assessments from early users complement the benchmark data, pointing to a significantly improved user experience. Reviewers noted that Sonnet 4.5 operates approximately 50% faster than its predecessors, citing an instance where it finished a code review in two minutes compared to ten minutes for GPT-5 Codex. The model also earns praise for being more reliable, easier to direct, and for its concise communication, which enhances its utility as a collaborative tool. The release was executed with a well-coordinated enterprise strategy, making the model immediately available on major platforms like Amazon Bedrock, Google Cloud Vertex AI, and GitHub Copilot to facilitate easy adoption. This rollout is supported by a new Claude Agent SDK and a competitive pricing structure that provides enhanced capabilities for the same price as the previous version, positioning Anthropic as a key AI provider for major cloud platforms.

Ethical alignment and safety are central to the model's design. Anthropic describes Sonnet 4.5 as its "most aligned frontier model," with significant decreases in undesirable behaviors such as deception and sycophancy, alongside improved defenses against security threats like prompt injection. This emphasis on safety is designed to appeal to enterprises, particularly those in regulated sectors. Looking ahead, the experimental "Imagine with Claude" feature provides a preview of Anthropic's long-term goals, demonstrating the model's ability to create interactive software interfaces in real-time and suggesting a future where the lines between using and developing software become blurred.

In conclusion, the launch of Claude Sonnet 4.5 heightens the competition within the AI industry, underscoring the rapid pace of development and the market's shift toward specialized models. By setting a new benchmark for performance in software development and enterprise automation, Anthropic has strategically positioned itself to secure a major role in this quickly advancing field.