anthropic launches claude sonnet 4.5, its best ai model for coding

Anthropic Introduces Claude Sonnet 4.5: A New Coding AI

Anthropic has unveiled its latest large language model, Claude Sonnet 4.5, on Monday. The company asserts this model achieves cutting-edge results in coding performance evaluations.

According to Anthropic, Claude Sonnet 4.5 is now capable of constructing fully functional applications, moving beyond the creation of mere prototypes. This signifies a substantial improvement in dependability compared to earlier AI iterations.

Availability and Pricing

Access to Claude Sonnet 4.5 will be provided through both the Claude API and the Claude chatbot interface.

Developers will find the pricing structure remains consistent with Claude Sonnet 4: $3 is charged per million input tokens – equivalent to approximately 750,000 words, exceeding the length of the complete “Lord of the Rings” saga – and $15 per million output tokens.

Growing Popularity and Competitive Landscape

Over the past year, Anthropic’s AI models have gained considerable traction among developers and businesses. This is largely attributable to their robust performance in software engineering applications.

Notably, both Apple and Meta are reported to be utilizing Claude AI models internally. Anthropic has also successfully generated revenue through API access, serving AI coding platforms like Cursor, Windsurf, and Replit.

However, OpenAI’s anticipated GPT-5 has recently presented a challenge to Anthropic’s leading position, demonstrating superior performance on various coding benchmarks.

Performance and Capabilities

Anthropic maintains that Claude Sonnet 4.5 delivers industry-leading performance on several coding benchmarks, including the SWE-Bench Verified assessment.

David Hershey, an AI researcher at Anthropic, explained to TechCrunch that benchmark scores alone don’t fully represent the model’s capabilities.

Hershey detailed instances where Claude Sonnet 4.5 autonomously coded for up to 30 hours during initial trials with enterprise clients. During this period, the AI not only developed an application but also established database services, acquired domain names, and conducted a SOC 2 audit to ensure product security.

Industry Reactions

Michael Truell, CEO of Cursor, stated in a release that Claude Sonnet 4.5 represents the current state-of-the-art in coding performance, particularly for extended tasks.

Jeff Wang, CEO of Windsurf, echoed this sentiment, describing Claude Sonnet 4.5 as ushering in “a new generation of coding models.”

Enhanced Safety and Alignment

Anthropic also asserts that Claude Sonnet 4.5 is its most aligned frontier AI model to date. It exhibits reduced tendencies towards sycophancy and deception compared to its predecessors.

Furthermore, the company has improved the model’s resistance to prompt injection attacks.

New Tools for Developers

Concurrent with the launch of Claude Sonnet 4.5, Anthropic is introducing the Claude Agent SDK. This SDK utilizes the same infrastructure as Claude Code, enabling developers to create their own intelligent agents.

A temporary research preview, “Imagine with Claude,” is also being released for Max subscribers. This preview demonstrates the AI generating software in real-time, responding to user requests without relying on pre-defined functions or existing code.

Rapid Innovation in the AI Sector

The competitive nature of the AI landscape has led to frequent releases of flagship models, often occurring every few months.

Claude Sonnet 4.5’s launch follows closely on the heels of Anthropic’s previous model, Claude Opus 4.1, which was released less than two months prior. This fast-paced development cycle makes it challenging for any single company to maintain a significant advantage for an extended period.