Anthropic has just presented Claude Sonnet 4.5, an evolution that the company defines as its most precise model to date. The focus is on agents , programming , and computer use , expanding on what they already offered in the previous versions of the Sonnet series. Its arrival is interpreted within an increasingly competitive landscape: OpenAI has launched GPT-5 with different levels of capability, and Google continues to bet on Gemini, creating an environment where each advancement generates new expectations.
The family’s trajectory helps to understand the place occupied by this new version. With Sonnet 3.7, Anthropic introduced a hybrid reasoning model that marked a significant leap in coding , content generation , and data analysis . The subsequent arrival of Sonnet 4 consolidated that leap, reinforcing its position as a practical option for users. These improvements made Sonnet into a notable alternative for programmers, and it is from this strong foundation that expectations rise regarding what 4.5 can contribute.
What Anthropic Promises with His New Model
Sonnet 4.5 introduces several improvements specifically designed for agents that require extended periods of attention. According to Anthropic, the model can maintain focus for more than 30 hours on complex tasks and is able to produce outputs of up to 64,000 tokens , significantly enhancing its capacity for planning and generating code in extensive blocks. Developers now have finer control over the time the “thinking” model spends before responding, allowing for a balance between speed and detail based on the specific needs of each project.
Another area where Sonnet 4.5 seeks to differentiate itself is in its performance within computer and browser environments. Anthropic points out that the model has achieved a score of 61.4% in Osworld , a benchmark that measures performance in completing real tasks within desktop scenarios. This substantial leap from the 42.2% achieved by Sonnet 4 just a few months ago showcases considerable progress. Practical examples of this advancement include demonstrations with its Chrome extension, where Claude can navigate websites, fill out spreadsheets, and conduct competitive analyses without requiring constant supervision.
In terms of programming , Sonnet 4.5 aims to solidify its leadership. Anthropic asserts that the model is capable of covering the entire software development cycle : from initial planning to the refactoring of large projects, as well as ongoing maintenance and error correction. With Claude Code’s support, it strives to become a reliable assistant for technical teams .
The applications of Sonnet 4.5 extend to various fields, making it a model designed for corporate and research environments . The frequently cited examples from its presentation include:
- Cybersecurity: Deployment of agents to rectify failures autonomously.
- Finance: Continuous monitoring of regulatory changes and proactive risk management.
- Productivity: Editing and creating office files in disparate formats.
- Investigation: Combining internal and external data to generate comprehensive reports.
- Content Creation: Writing that incorporates mathematical understanding and nuanced semantic analysis.
Moreover, the company states that Sonnet 4.5 has undergone evaluations by external experts to ensure its safety and reliability .
Sonnet 4.5 is now accessible to all users on Claude.AI, available both on the web and through applications for iOS and Android. Developers can integrate it into the Claude Developer Platform, as well as services like Amazon Bedrock and Google Cloud Vertex AI. The free plan operates with a session limit that resets every five hours, alongside a variable number of messages according to demand. Pricing starts at $3 per million input tokens and $15 per million output tokens.
Images | Anthropic | Xataka with Gemini 2.5
In Xataka | “The humanoid robots are pure fantasy”: Irobot’s co-founder believes that there is a robotics bubble.