Anthropic today announced the launch of Claude 3.5 Sonnet, the first product in the Claude 3.5 series. This model outperforms its competitors and its predecessor, Claude 3 Opus, in multiple evaluations, while maintaining speed and cost comparable to mid-range models, setting a new industry standard.
Claude 3.5 Sonnet is now available to the public on Claude.ai and the Claude iOS app, and is also offered commercially through the Anthropic API, Amazon Bedrock, and Google Cloud's Vertex AI. The model charges $3 per million input tokens and $15 per million output tokens, with a context window of 200K tokens.
Significant Performance Enhancements
Claude 3.5 Sonnet sets new industry benchmarks in graduate-level reasoning, undergraduate-level knowledge, and coding abilities. It has made significant strides in understanding nuances, humor, and complex instructions, and can create high-quality content in a natural and engaging tone.
Notably, Claude 3.5 Sonnet operates at twice the speed of Claude 3 Opus. In internal agent coding evaluations, Claude 3.5 Sonnet solved 64% of the problems, far exceeding Claude 3 Opus's 38%. This makes it particularly suitable for handling complex tasks such as context-sensitive customer support and multi-step workflow coordination.
Significant Enhancement in Visual Capabilities
Claude 3.5 Sonnet has also made major breakthroughs in visual processing, surpassing Claude 3 Opus in standard visual benchmark tests. It excels in tasks requiring visual reasoning, such as interpreting charts and graphs, and can accurately transcribe text from imperfect images, which is of great significance to industries such as retail, logistics, and financial services.
New Feature: Artifacts
Anthropic has also introduced the Artifacts feature on Claude.ai, expanding the ways users can interact with Claude. Users can ask Claude to generate code snippets, text documents, or website designs, which will be displayed in a dedicated window alongside the conversation, creating a dynamic workspace.
Commitment to Safety and Privacy
Despite the leap in intelligence of Claude 3.5 Sonnet, Anthropic states that it still maintains an ASL-2 safety level. The company has collaborated with external experts, including the UK Artificial Intelligence Safety Institute (UK AISI), to conduct rigorous safety tests on the model. Anthropic emphasizes that it will not use user-submitted data to train its generative models unless explicitly permitted by the user.
Future Outlook
Anthropic plans to release Claude 3.5 Haiku and Claude 3.5 Opus later this year to complete the Claude 3.5 series. The company is also developing new modes and features to support more enterprise use cases, including integration with enterprise applications and personalized memory functions.
Anthropic invites users to submit feedback directly within the product to help improve Claude 3.5 Sonnet and guide future development paths.