Source: aboutamazon.com/news
In the public beta of Anthropic’s Claude 3.5 Sonnet, the ability to generate computer actions such as keystrokes and mouse clicks to perform tasks via user interfaces has been introduced. Also, artificial intelligence has received significant improvements, in particular in writing software code.
The new version of the Claude 3.5 Sonnet, the newest and most intelligent model in the Claude family of models from security and artificial intelligence (AI) research company Anthropic, is now available on Amazon Bedrock. This version offers ground-breaking capabilities with improved AI agent-computer interactions. Amazon Bedrock offers the broadest selection of high-performance base models (FM) from leading AI companies such as Anthropic, along with the capabilities and enterprise security customers need to rapidly build and deploy generative AI applications.
Anthropic‘s data shows that the Claude 3.5 Sonnet model delivers overall improvements over its predecessor, with particularly significant gains in encoding, an area where it was already a leader. It improves SWE-bench Verified performance from 33.4% to 49%, scoring higher than all public models, and excels in computer vision and writing tasks. The enhanced reasoning and problem-solving skills of the Claude 3.5 Sonnet make it an ideal model for complex cognitive tasks such as understanding subtle instructions and context, generating creative solutions to problems, recognizing and correcting errors, and performing complex data analysis.
The multimodal Claude 3.5 Sonnet also excels in image processing with industry-leading computer vision, especially when interpreting charts and graphs, helping you analyze data faster and more deeply. It can accurately decode text from imperfect images, such as poorly scanned documents, and thus extract as much information as possible. The improved Claude 3.5 Sonnet model offers these improvements at the same speed as the original Claude 3.5 Sonnet model, and at no additional cost.
Additionally, an experimental computer capability (in public beta) allows Claude 3.5 Sonnet to generate computer actions such as keystrokes and mouse clicks to perform tasks with user interfaces (UIs) that may require dozens of even hundreds of steps. This capability can be used by developers to perform routine tasks such as online research and expense logging, creating promising use cases from software testing to automated research. According to Anthropic, this is the first advanced AI model to use computers in this way. Anthropic expects these capabilities to improve over time.
Benefits of the updated Claude 3.5 Sonnet include:
- Advanced coding capabilities: High performance in planning and solving complex coding tasks such as code migration, code patching, and translation into another programming language.
- Enhanced context understanding: The model combines deep context awareness, advanced logic, and synthesis to deliver extremely fast responses and natural, human-like interactions for a variety of use cases such as content creation, document Q&A, and customer service.
- Agent Tasks and Tool Use: The model demonstrates high performance in instruction execution, tool selection, error correction, and advanced logic for agent workflows that require tool use.
Coming soon is Claude 3.5 Haiku, the latest generation of Anthropic’s fastest model, which will initially be available in text-only format, with image support to follow. This model is optimal for scenarios where speed and availability are important, such as chatbots, code completion, data extraction and labeling, and processing large amounts of information such as purchase history, pricing, or inventory data.
AWS customers across industries already have access to Claude’s advanced models through Amazon Bedrock, enabling them to rapidly test, build, and deploy generative AI applications in their organizations.
- Improving communication with customers. DoorDash, a technology company that connects consumers with their favorite local businesses in more than 30 countries around the world, receives hundreds of thousands of support requests through its contact center every day. The company integrated Anthropic’s Claude models through Amazon Bedrock with Amazon Connect and Amazon Lex, creating a generative AI solution for contact centers that optimizes customer service. Using Knowledge Bases for Amazon Bedrock enabled DoorDash to reduce development time for generative AI applications by 50%. This solution handles hundreds of thousands of calls every day, reducing call volume and reducing response time to 2.5 seconds or less (by 50%). It has also reduced thousands of escalations to operators every day, resulting in faster delivery and increased customer satisfaction.
- Strengthening human creativity. WPP, the world’s largest advertising company, has integrated Claude models through Amazon Bedrock into its intelligent marketing operating system, WPP Open. This enables WPP’s 114,000 employees around the world to improve the quality of work for clients and increase the efficiency of tasks, from idea generation and content creation to copywriting and design.
- Assistance to people with learning disabilities. Swindon Borough Council (UK) is using generative AI to adapt complex documents into a more accessible format for people with disabilities. ‘Simply Readable’ is based on the work the new technology team has done with volunteers to better understand the barriers faced by people with disabilities. The solution includes features such as large font size, increased spacing between words and the ability to use images to facilitate comprehension. Swindon Council has now made an updated solution created with the help of Claude at Amazon Bedrock. It is free worldwide so that the technology can help educate as many people with disabilities as possible, empowering them to make informed decisions in key areas of their lives.
The availability of Claude 3.5 models on Amazon Bedrock confirms Anthropic’s long-term commitment to providing AWS customers with access to future generations of their FMs through Amazon Bedrock. With new industry-leading models such as the Claude 3.5 Sonnet and Claude 3.5 Haiku, Amazon Bedrock customers can choose the models that best suit their industries and use scenarios, regardless of price, performance or feature requirements. Customers benefit from Amazon Bedrock’s enterprise-grade security and privacy, combined with Anthropic’s built-in approach to responsible and secure AI model development. With Amazon Bedrock’s simple and unified API, customers can easily leverage the power of the underlying models and manage integrations, enabling rapid innovation as more performant and accessible models are deployed.
We are happy to announce that the Wise IT team has upgraded AWS partner status to Advanced! This achievement confirms the competence of our engineers in cloud migration and implementation of AWS solutions in customer infrastructure. Looking to implement AWS solutions into your infrastructure? Get a free consultation today: