Anthropic releases Claude 3.5 Sonnet and Artifacts

Background

Claude 3.5 Sonnet was the first model of the Claude 3.5 family, which Anthropic introduced with this release.^[1][2]

What happened

On 20 June 2024 Anthropic released Claude 3.5 Sonnet. It was the first model of its Claude 3.5 family.^[1][2]

Claude.ai also gained Artifacts. Artifacts is a window beside the chat where generated code, documents and website designs appear, and where they can be edited.^[1]

Key details

The model cost $3 per million input tokens and $15 per million output tokens, the same as Claude 3 Sonnet. It had a 200K-token context window and ran at twice the speed of Claude 3 Opus.^[1]

It was free on claude.ai and the Claude iOS app, and it was available through the Anthropic API, Amazon Bedrock and Google Cloud Vertex AI.^[1]

The UK AI Safety Institute carried out pre-deployment testing of the model. Anthropic kept the model at AI Safety Level 2 (ASL-2).^[1]

Anthropic said that Claude 3.5 Haiku would follow later in the year.^[1]

Impact

The release showed a mid-tier price point carrying a model that ran twice as fast as Claude 3 Opus. Artifacts gave users a window beside the chat for generated code, documents and designs.^[1]

Agentic coding refers to a model carrying out multi-step programming work on its own, rather than only answering a single question. Running at the mid-tier price meant users could get faster results without paying for the top model.

The UK AI Safety Institute is a government body that tests AI models for safety risks. Pre-deployment testing means the institute examined the model before it was released to the public.

Artifacts matter because they separate the output from the chat. A user can read or edit a generated document or piece of code in its own window, and keep the conversation for questions and changes.