Claude Sonnet 4.5 is live! Anthropic: Directly capable of launching products.

Anthropic officially released the latest generation programming model Claude Sonnet 4.5 on 9/29, emphasizing that this new model can directly generate "products that can be officially launched" while maintaining the same API pricing standards as the previous generation.

It can produce software that goes online directly, with the same charging standard as the previous generation.

Anthropic was the first to point out that Claude Sonnet 4.5 can directly produce "deployable applications," and the pricing is the same as the previous generation Sonnet 4, which is $3 per million input tokens and $15 per million output tokens. This translates to processing about 750,000 words, a length that even exceeds the total length of The Lord of the Rings trilogy.

Can automatically code for 30 hours continuously, with significantly improved security.

Anthropic researcher David Hershey emphasized during an interview that the capabilities of Claude Sonnet 4.5 go far beyond what test numbers can demonstrate. He witnessed firsthand during early trials at the company that this model could automatically write code for 30 consecutive hours, not only independently completing application architecture but also building database services, purchasing domain names, and even conducting SOC 2 security audits to ensure the product meets safety standards.

The official added that Claude Sonnet 4.5 is currently the most stable and reliable flagship model, significantly reducing sycophantic responses and misleading behavior, and is also more effective at resisting malicious prompt attacks.

Claude Agent SDK and Imagine with Claude new features launched simultaneously.

In addition to the main model, Anthropic has launched two new features simultaneously. The first is the Claude Agent SDK, which allows developers to build their own AI assistants using infrastructure equivalent to Claude Code.

The second is Imagine with Claude, which is a research version feature provided for subscribers of the highest paid plan (Max). It can instantly generate software, respond to user needs, and has no pre-written functions or codes; it is a real-time programming assistant for users.

The competition among AI models is becoming intense, making it difficult to maintain a leading advantage.

Currently, the competition in AI is fierce, and companies are launching flagship models at an increasingly rapid pace. Anthropic's previous flagship, Claude Opus 4.1, was only released less than two months ago. Recently, OpenAI's GPT-5 has also surpassed the Claude series in several programming assessments, challenging Anthropic's advantage.

Against this backdrop, the rapid launch of Claude Sonnet 4.5 demonstrates that the competitive advantage of AI models is difficult to maintain for an extended period.

(The latest five mainstream AI language models (LLM) fully analyzed, with paid options, applications, and security all understood at once)

This article Claude Sonnet 4.5 is online! Anthropic: can directly produce online products first appeared in Chain News ABMedia.