HomeWinBuzzer NewsAnthropic´s New Claude 3.5 Sonnet AI Model Beats OpenAI´s GTP-4o in Benchmarks

Anthropic´s New Claude 3.5 Sonnet AI Model Beats OpenAI´s GTP-4o in Benchmarks

The Model shows strong results across AI assessments, including tasks in reading, coding, math, and visual processing, and vision-related assignments.

-

Anthropic has introduced its newest AI model, Claude 3.5 Sonnet, promising increased performance and efficiency. The update, designed to analyze both text and images and generate text, is a marked upgrade over its predecessors, Claude 3 Sonnet and Claude 3 Opus.

Anthropic’s latest model emphasizes incremental progress, with Claude 3.5 Sonnet operating at approximately double the speed of Claude 3 Opus. This makes it particularly useful for applications needing rapid responses, such as customer service chatbots. The context window remains at 200,000 tokens, around 150,000 words, allowing the model to manage extensive text analyses.

Enhanced Capabilities and Benchmarks

Claude 3.5 Sonnet achieves strong results across AI assessments, including tasks in reading, coding, math, and visual processing. The model has outshone major competitors, including OpenAI’s GPT-4o, in various evaluations.

During an internal coding assessment, Claude 3.5 Sonnet resolved 64% of the issues, surpassing Claude 3 Opus, which resolved 38%. The evaluation measured the model’s proficiency in fixing bugs or enhancing functionality within an open-source codebase based on a natural language description of the required improvements.
 
Anthropic Claude 3.5 Sonnet Benchmarks official

Once directed and equipped with Anthropic´s own tools, Claude 3.5 Sonnet autonomously writes, modifies, and runs code, demonstrating advanced reasoning and problem-solving skills. Its adeptness at code translation makes it especially valuable for modernizing legacy systems and transitioning codebases.

These enhancements are tied to updates in architecture and the integration of novel training data, although details on the data sets have not been disclosed.

Vision Task Improvements

Claude 3.5 Sonnet also shows substantial progress in vision-related assignments. The model now interprets charts and graphs with better accuracy and can transcribe text from distorted or low-quality images. The improvements are set to boost application performance where quick and precise visual data analysis is required.
 

New Artifacts Workspace for Content Creation

Coinciding with the new model’s release is Artifacts, a workspace feature geared towards editing and enhancing AI-generated content like code and documents. Currently in preview, Artifacts will soon include team collaboration features and knowledge base storage, simplifying content development and refinement.

When users request Claude to produce content such as code snippets, text documents, or website designs, these artifacts are displayed in a dedicated window next to their conversation. This feature creates a dynamic workspace where users can view, modify, and expand upon Claude’s output in real-time, allowing for the seamless integration of AI-generated content into their projects and workflows. This functionality signifies Claude’s transition from a conversational AI to a collaborative work environment.
  

Commitment to Safety and Privacy

Anthropic says that its latest AI model, Claude 3.5 Sonnet, has been rigorously tested to prevent misuse. Despite its advanced capabilities, the model maintains an ASL-2 safety level.
 
AI Safetz Levels ASL-1 ASL-2 ASL-3 ASL-4 via Anthropic

The company emphasized its commitment to safety and transparency, highlighting collaboration with external experts. Claude 3.5 Sonnet was evaluated by the UK’s Artificial Intelligence Safety Institute (UK AISI) before deployment, and its findings were shared with the US AI Safety Institute under a partnership agreement.

Anthropic also incorporated feedback from external experts to ensure robust evaluations and address emerging abuse trends. Input from child safety experts at Thorn helped update classifiers and fine-tune the model.

Anthropic stresses that privacy remains a core principle in AI development. It promises not to use user-submitted data for training without explicit permission and that it has not used any customer data in its generative models to date.

Availability and Access

Claude 3.5 Sonnet is accessible to free users of Anthropic’s web client and the Claude iOS app. Subscribers on paid plans are granted higher usage limits. Additionally, the model is available through Anthropic’s API and managed platforms like Amazon Bedrock and Google Cloud’s Vertex AI.

Anthropic focuses on building an ecosystem around its models, investing in tools such as the experimental steering AI, and expanding product availability.

SourceAnthropic
Markus Kasanmascheff
Markus Kasanmascheff
Markus has been covering the tech industry for more than 15 years. He is holding a Master´s degree in International Economics and is the founder and managing editor of Winbuzzer.com.

Recent News

Mastodon