Anthropic releases new AI model that can communicate with any desktop app

Share Now

Anthropic recently launched an enhanced iteration of Claude 3.5 Sonnet, now equipped with a Computer Use functionality that categorizes prompts into user commands suitable for desktop operations. Additionally, the introduction of the new Claude 3.5 Haiku AI model has been announced by the company.

Briefly speaking

Anthropic claims that Claude 3.5 Sonnet surpasses OpenAI o1-preview in coding prowess.
The functionality of the Computer Use feature mimics human interaction with a computer, operating in a manner akin to a person using a PC.
Introducing the Claude 3.5 Haiku model, renowned for its exceptional efficiency while maintaining an affordable price point within the series.

Anthropic has recently revealed its latest AI models – an upgraded version of Claude 3.5 Sonnet and a new iteration called Claude 3.5 Haiku. The improved Claude 3.5 Sonnet demonstrates enhancements in all aspects compared to its predecessor, showcasing significant progress in coding, an area where it previously excelled. On the other hand, Claude 3.5 Haiku delivers performance on par with the previous flagship model, Claude 3 Opus, across various evaluations, while maintaining the same affordability and speed as its predecessor. The standout feature is the introduction of the Computer Use functionality, a notable addition introduced during the public beta phase of the Claude 3.5 Sonnet model. Let’s delve into a comprehensive overview of all the new components.

Anthropic’s latest report highlights significant enhancements in industry standards within the revised Claude 3.5 Sonnet, showcasing notable advancements in agentic programming and tool manipulation assignments. The enhanced model now boasts the capability to execute a diverse range of desktop functions, enabling seamless web browsing and utilization of various desktop applications by Claude 3.5 Sonnet.

The issue of AI dominance is a prominent concern in this context. The company has reassured that humans will retain control. According to TechCrunch, users will oversee all tasks by offering explicit prompts that guide Claude’s actions. Through the Computer Use feature, user prompts are deconstructed into computer commands for accomplishing specific tasks.

The company asserts that the enhanced iteration represents a superior and more robust model. It contends that the 3.5 Sonnet outperformed other well-known models such as OpenAI o1 in terms of coding performance.

The company revealed in its statement that the enhanced Claude 3.5 Sonnet has garnered positive reviews from early users, showcasing a notable advancement in AI-driven programming. Developers can now utilize Computer Use via Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI platform. The latest iteration of the 3.5 Sonnet, which excludes Computer Use, is now integrated into Claude apps, delivering multiple performance upgrades over its predecessor.

Anthropic recognizes that the enhanced 3.5 Sonnet still falls short of perfection, manifesting challenges in basic functions such as scrolling and zooming. Additionally, it may overlook fleeting interactions and alerts as a result of its screenshot capturing and arranging approach. “Claude’s Computer Use continues to exhibit sluggish performance and frequent errors,” notes Anthropic in its publication. The company advocates for developers to initiate their testing with less critical operations.

Haiku Trio by Claude

Anthropic has revealed exciting news about the imminent release of an enhanced edition of Haiku, the most budget-friendly and effective model in its Claude series. The upcoming Claude 3.5 Haiku, scheduled for launch in the forthcoming weeks, will provide comparable performance to the previously top-tier model Claude 3 Opus in certain metrics. This achievement will be accomplished while keeping the price unchanged and retaining the “similar speed” characteristic of Claude 3 Haiku.

Anthropic highlighted in a blog post that Claude 3.5 Haiku is ideal for front-end products, niche sub-agent duties, and crafting tailored experiences by leveraging extensive data sets such as purchase records, pricing information, and inventory data, thanks to its low latency, enhanced instruction tracking, and precise tool utilization.