Anthropic unveils Claude 3.5 Sonnet, its most performant AI model yet

Anthropic, a leading AI research company, has released Claude 3.5 Sonnet, setting new standards for AI in reasoning, learning, coding, visual processing and even humor understanding.

Claude 3.5 Sonnet (source: Anthropic)

Claude 3.5 Sonnet is available for free on and the Claude iOS app. Subscribers to the Claude Pro and Team plans have access to higher rate limits. Additionally, the model is accessible via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. The model is priced at $3 per million input tokens and $15 per million output tokens, offering a context window of 200,000 tokens.


Claude 3.5 Sonnet outperforms Claude 3 Opus and competitor models such as GPT-4o, Gemini 1.5 Pro, and Meta’s Llama-400B in various benchmarks including reasoning and knowledge tasks, object detection and image classification. Additionally, its enhanced capability to both create code and comprehend existing code makes it valuable for software development tasks. More results can be seen in the table below.

Claude 3.5 overall benchmark scores (source: Anthropic)

According to Anthropic, Claude 3.5 Sonnet is the strongest vision model in its series, outperforming Claude 3 Opus on standard vision benchmarks. The improvements are particularly significant in tasks that require visual reasoning, such as interpreting charts and graphs. Additionally, Claude 3.5 Sonnet has the ability to accurately transcribe text from imperfect images.

Claude 3.5 vision benchmark scores (source: Anthropic)
Claude 3.5 Sonnet for vision (source)

Claude 3.5 Sonnet’s key improvements

  • Visual processing: It can interpret charts, graphs, and even extract text from images.
  • Natural language generation: The model now understands humor and nuance, allowing it to create high-quality, engaging content with a more natural tone.
  • Customer support: Claude 3.5 Sonnet is able to perform complex inquiries and multi-step workflows, analyze unstructured data, generate insights, and create visualizations, providing a more natural and efficient customer support experience.
  • Coding: The model can write, edit, and even execute code.

Artifacts—a new way to interact with Claude has introduced a new feature called Artifacts, transforming the platform from a chat-based AI into a collaborative workspace. Users can now generate and interact with substantial content like code snippets, text documents, or designs in a separate window from the main conversation, allowing for real-time editing and integration into projects. This addition is a move towards a future where will support team collaboration, enabling groups and organizations to work with Claude as a virtual team member.

Anthropic plans to release new versions, Claude 3.5 Haiku and Claude 3.5 Opus, later in the year to complete the Claude 3.5 series. They’re also developing new features and integrations for business applications, including a memory feature for a more personalized user experience.

Read more:

Announcement release on Anthropic: “Claude 3.5 Sonnet”

Other popular posts