Nov. 4, 2024: This article has been updated to reflect the general availability of Anthropic’s Claude 3.5 Haiku.
October 22, 2024: This article has been updated to reflect the general availability of Anthropic’s upgraded Claude 3.5 Sonnet in Amazon Bedrock with a new computer use capability available in public beta.
Claude 3.5 Haiku and Claude 3.5 Sonnet, the latest and most intelligent models in the Claude model family from artificial intelligence (AI) safety and research company Anthropic, are now generally available in Amazon Bedrock. The upgraded Claude 3.5 Sonnet also includes a groundbreaking new computer use capability in public beta. Amazon Bedrock provides the broadest selection of high-performing foundation models (FMs) from leading AI companies, like Anthropic, along with the capabilities and enterprise security customers need to quickly build and deploy generative AI applications.
Data from Anthropic shows that Claude 3.5 Sonnet delivers across-the-board improvements over its predecessor, with particularly significant gains in coding—an area where it already led the field. It improves performance on SWE-bench Verified from 33.4% to 49%, scoring higher than all publicly available models, while also excelling at vision and writing tasks. Claude 3.5 Sonnet’s enhanced reasoning and problem-solving skills makes it the ideal model for complex cognitive tasks such as understanding nuanced instructions and context, generating creative solutions to problems, recognizing and correcting mistakes, and delivering more sophisticated analysis around complex data.
The multimodal Claude 3.5 Sonnet also excels at processing images with industry-leading vision, particularly when interpreting charts and graphs—helping to get faster, deeper insights from data. It can accurately decipher text from imperfect images—for example, poorly scanned documents—and in doing so, glean more insights than from text alone. The enhanced Claude 3.5 Sonnet model offers these advancements at the same speed as the original Claude 3.5 Sonnet model and without additional cost.
Additionally, a groundbreaking experimental computer use capability (in public beta) enables Claude 3.5 Sonnet to generate computer actions—like keystrokes and mouse clicks—to accomplish tasks using user interfaces (UI) that can require dozens, even hundreds, of steps. The capability can be used by developers to handle routine tasks like online research and expense filing, driving promising use cases from software testing to automated research, while balancing safe and responsible deployment. According to Anthropic, it is the first frontier AI model to use computers in this way, albeit experimentally, and Anthropic expects the capability to improve over time.
Upgraded Claude 3.5 Sonnet strengths
- Advanced coding capabilities: Offering strong performance in planning and solving for complex coding tasks, ranging from code migrations, code fixes, and translations.
- Improved understanding of context: Combining strong context comprehension, advanced reasoning, and synthesis to deliver impressively fast response times and natural-sounding, human-like interactions across use cases like content generation, document Q&A, and customer service.
- Agentic tasks and tool use: Excelling in instruction following, tool selection, error correction, and advanced reasoning for agentic workflows that require the use of tools.
Finally, Claude 3.5 Haiku, the latest generation of Anthropic’s fastest model, is now available as a text-only model with image to follow and is optimal for use cases where speed and affordability matter. Claude 3.5 Haiku improves across every skill set and surpasses even Claude 3 Opus, Anthropic’s prior largest model, on many intelligence benchmarks, and at speeds similar to the previous generation of Haiku.
Claude 3.5 Haiku strengths
- Code completions: Providing quick, accurate code suggestions and completions in real-time development workflows as a result of its rapid response time and understanding of programming patterns.
- Interactive chatbots: Creating responsive, engaging chatbots that can handle high volumes of user interactions efficiently.
- Data extraction and labeling: Leveraging its improved analysis skills to efficiently process and categorize data.
- Data processing: Synthesizing huge volumes of data—like purchase history, pricing, or inventory data—to generate personalized experiences.
AWS customers across a range of industries have already been accessing the state-of-the-art Claude models through Amazon Bedrock, enhancing their ability to rapidly test, build, and deploy generative AI applications across their organizations.
- Improving customer connections. DoorDash, a technology company that connects consumers with their favorite local businesses in more than 30 countries worldwide, receives hundreds of thousands of requests for assistance through its contact center each day. The company has incorporated Anthropic's Claude models in Amazon Bedrock with Amazon Connect and Amazon Lex to build a generative AI contact center solution to streamline customer support. By deploying Knowledge Bases for Amazon Bedrock as the foundation, DoorDash reduced generative AI application development time by 50%. The solution is currently fielding hundreds of thousands of support calls each day and has driven material reductions in call volumes, while trimming response time to 2.5 seconds or less (a 50% reduction). The solution has led to a decrease in the number of escalations to customer service representatives by thousands per day, ultimately accelerating delivery time and increasing satisfaction.
- Enhancing human creativity. WPP, the world’s largest advertising company, has integrated Claude models using Amazon Bedrock into WPP Open, its intelligent marketing operating system. This enables WPP’s 114,000 employees worldwide to enhance the quality of client work, and improve the efficiency of tasks ranging from ideation and content generation, to copywriting and design.
- Helping people with learning disabilities. Swindon Borough Council, a UK local authority, is using generative AI to convert complex documents into a more accessible format for people with learning disabilities. ‘Simply Readable’ is based on work the council’s emerging technology team undertook with volunteers, to better understand some of the obstacles they face in their lives. It incorporates features such as large font size, increased spacing between words, and the option to have images to help with comprehension. The council has now made the solution—built using Claude in Amazon Bedrock— license-free, worldwide, so the technology can benefit as many people with learning disabilities as possible, so they can make informed decisions about key areas of their lives.
The availability of Claude 3.5 models in Amazon Bedrock delivers on Anthropic’s long-term commitment to provide AWS customers with access to future generations of its FMs via Amazon Bedrock. As new, industry-leading models like Claude 3.5 Sonnet and Claude 3.5 Haiku are introduced, customers can use Amazon Bedrock to access and choose among models that best suit their industries and use cases, whether price, performance, or capability requirements. Customers benefit from the enterprise-grade security and privacy of Amazon Bedrock, coupled with Anthropic’s built-in approach to responsibly and safely develop AI models. Using Amazon Bedrock’s simple and unified API, customers can seamlessly harness the power of foundation models and manage integrations, enabling rapid innovation as more performant and affordable models are launched.
Learn more about Amazon Bedrock and check out additional info on the AWS News blog about Claude 3.5 Haiku and Claude 3.5 Sonnet.
