Follow ZDNET: Add us as a favorite source On Google.
Key takeaways of zdnet
- The latest model of anthropic, Cloud Sonnet 4.5, is here.
- It scored a lot on the coding benchmark.
- The cloud code also received a long -awaited upgrade.
Anthropic’s coding tools are well considered among developers, with its cloud 4 sonnet model, released in May, serving as an independent and reliable coding assistant for many. A few months later, Anthropic released its next generation model, characterizing the upgradation of its performance across the board.
Also: Anthropic’s Free Cloud Sonnet 4 handled my coding tests – but it did not pay it in any way.
Cloud Sonnet 4.5 is now available, the company said on Wednesday, claiming that it is “the best coding model” in the world as well as the best model to build complex agents, use computers and get logic and mathematical abilities. Anthropic also launched an update about its cloud code, cloud for chrome extension, and more.
What’s new in Cloud 4.5 Sonnet?
If you are closely tracking the release of anthropic, you can remember that in May, Cloud Opus 4 and Sonate 4 scored the highest score among the frontier models on the Industry-standard software engineering benchmark test (SWE-Bench), which evaluates llms capables to solve the real-ward software engineering functions from GITHUB. Cloud Opus 4.1 released in August crossed it.
Now, Cloud Sonnet 4.5 has applied the last model that performs it better on the Swe-Bench verified evaluation, a human-filter of Swe-Bench. Cloud Sonnet 4.5 also improved the leading model from contestants including GPT-5 Codex, GPT-5, and Gemini 2.5 Pro.
Anthropic stated that on SWE-Bench verified, Sonnet 4.5 focused its attention for more than 30 hours on complex, multi-step tasks. This capacity is especially useful for agentic functions, which often requires single function in the background for an extended period.
Also: I got 4 years product development in 4 days for $ 200, and I am still shocked
Other reforms include its performance on computers, as is indicated by its score on the Osworld Benchmark, which tests the performance of the AI ​​model on real -world computer tasks:
Anthropic said in release, “Sonnet 4.5 is now 61.4%. Four months ago, Sonnet 4 gained 42.2%.”
Cloud for Chrome extension, now rolled out for all those who join the weightlist last month, uses these abilities. Anthropic also saw improvement in mathematics and logic.
Also: I prepared two AI tools to solve a major bug – but they could not do it without me
According to the company, Cloud Sonnet 4.5 is also the “most coalition” frontier model of anthropic. This means that this is the model that follows humans’ instructions and the most closely of matters of intended use, and this has reduced examples of behavior such as sycophancy and deception. The model is also better in opposing the early injection attacks and has AI Safety 3 (ASL -3) security on anthropic’s model framework.
How to access
Cloud Sonnet 4.5 is available everywhere, including Cloud.AI Chatbot. Of course, developers and professional can use new models in APIs and cloud codes, and for the same price of sonnet 4.
Other upgrades
Anthropic also upgraded its other coding prasad, which begins with the cloud code, now there are checkpoints that allow users to save progress and re -look at the previous state. It is also said that the anthropic is calling a “fresh” terminal interface and a native versus code extension.
Also: Openai, anthropropic, meta, and google released researchers joint AI Safety Warning – Why is here
Anthropic also launched Cloud Agent SDK, which is the same infrastructure that powers the cloud code, allowing developers to build their own agents with it.
Cloud Code API has introduced a new reference editing facility and a memory tool that, according to the company, enables agents to work more efficiently and deal with more complex problems. The company also upgraded the cloud apps to execute the code and make files in the chat.
