Claude 3.5 Sonnet

Anthropic has released Claude 3.5 Sonnet, the first in its upcoming Claude 3.5 model family. This new model surpasses industry benchmarks in various areas, including graduate-level reasoning, undergraduate-level knowledge, and coding proficiency. Claude 3.5 Sonnet operates at twice the speed of Claude 3 Opus while maintaining cost-effectiveness, making it ideal for complex tasks like customer support and multi-step workflows.

Key Features of Claude 3.5 Sonnet:

• Enhanced Intelligence: Outperforms competitor models and previous Claude iterations in various evaluations.
• Improved Coding Proficiency: Solves 64% of problems in an agentic coding evaluation, significantly exceeding Claude 3 Opus's 38% success rate.
• Advanced Vision Capabilities: Surpasses Claude 3 Opus in standard vision benchmarks, particularly excelling in tasks requiring visual reasoning, such as interpreting charts and graphs.
• Artifacts Feature: Introduces a new way to interact with Claude, enabling users to directly view, edit, and build upon generated content in real-time.

Availability and Pricing:

• Free access on Claude.ai and the Claude iOS app.
• Increased rate limits for Claude Pro and Team plan subscribers.
• Available via Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI.
• Costs $3 per million input tokens and $15 per million output tokens, with a 200K token context window.

Commitment to Safety and Privacy:

• Rigorous testing and training to mitigate misuse.
• External expert review and pre-deployment safety evaluation by the UK Artificial Intelligence Safety Institute (UK AISI).
• Integration of policy feedback from subject matter experts to enhance safety assessments.
• No user-submitted data is used to train generative models unless explicit permission is granted.

Future Plans:

• Release of Claude 3.5 Haiku and Claude 3.5 Opus later this year.
• Development of new modalities and features to expand business use cases.
• Exploration of features like Memory to personalize and enhance user experience.