Anthropic recently announced its new AI models – an updated Cloud 3.5 Sonnet and the new Cloud 3.5 Haiku. The upgraded Cloud 3.5 Sonnet offers improvements in all areas compared to its predecessor, with notable advancements especially in coding, a domain where it was already a leader. The Cloud 3.5 Haiku achieves performance equivalent to the previous largest model, the Cloud 3 Opus, in several assessments, while maintaining the same cost and speed as the previous Haiku version. But the main attraction here is the new Computer Usage feature, a notable addition to the public beta phase of the Cloud 3.5 Sonet model. Let’s take a detailed look at all the new entities.

Cloud 3.5 Sonnet with computer usage

Anthropic said the updated Cloud 3.5 Sonnet shows broad improvements on industry benchmarks, with particularly strong gains in agentic coding and tool-use tasks. This model can now execute any desktop-level command. This means that with the ability to browse the web, the Cloud 3.5 Sonnet can access any website and any application on the desktop.

However, concerns about AI taking over have increased here. On this the company has said that the human remains in control. TechCrunch reported that users would complete all tasks by providing specific prompts that would give a path to the cloud’s actions. A computer usage feature breaks down user signals into computer commands to achieve a certain task.

According to the company, the updated version is a better and stronger model. It claims that the 3.5 Sonnet has scored higher in coding performance than other popular models including OpenAI o1.

In the announcement, the company shared, “Early customer feedback shows that Advanced Cloud 3.5 Sonnet represents a significant leap forward for AI-powered coding.” Developers can access compute usage through Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI platform. The updated 3.5 Sonnet, being introduced for cloud apps without computer usage, offers several performance enhancements compared to the previous 3.5 Sonnet model.

But perfection is yet to come. Anthropic admits that the improved 3.5 Sonnet has difficulty with simple tasks like scrolling and zooming, and that “ephemeral” actions and notifications may be missed due to the method used to capture and assemble screenshots. “Computer use of the cloud remains slow and often error-prone,” Anthropic writes in his post. “We encourage developers to begin exploration with lower-risk tasks.”

cloud 3 haiku

Anthropic announced that an updated version of Haiku, the most affordable and efficient model of its cloud series, is coming soon. The upcoming Cloud 3.5 Haiku, which is scheduled to launch in the next few weeks, will deliver performance equivalent to Cloud 3 Opus, previously Anthropic’s most advanced model, on specific benchmarks, while maintaining the same cost and “same speed” as Cloud 3 Haiku. Will keep.

“With lower latency, better instruction following and more accurate tool usage, Cloud 3.5 Haiku is suitable for user-facing products, specialized sub-agent functions and generating personalized experiences from large amounts of data – such as purchase history, pricing , or inventory data, Anthropic wrote in a blog post.

Cloud 3.5 Haiku will initially be available as a text-only model and later as part of a multimodal package that can analyze both text and images.