SAN FRANCISCO — The AI foundation model company Anthropic is rolling out upgrades to its Claude family of models.
Anthropic is introducing its “computer use” feature as well as upgrades to Claude 3.5 Sonnet and the release of Claude 3.5 Haiku, according to the company today.
AI 'Computer Use'
The API-based "computer use" feature is designed to automate operations by performing computer tasks “the way people do." Developers can use the feature to direct Claude to look at a screen, move a cursor, click buttons and type text.
Anthropic says the "still-experimental" feature is sometimes "cumbersome and error-prone."
"We're releasing computer use early for feedback from developers and expect the capability to improve rapidly over time," the company says.
Several companies are using the feature, such as Asana, Canva, and DoorDash, and completing tasks that require dozens to hundreds of steps.
The feature is in public beta and currently available with Claude 3.5 Sonnet.
Claude 3.5 Sonnet
Anthropic reports the upgraded Claude 3.5 Sonnet — a "hardworking," mid-tier model — demonstrates improvements on several benchmarks, particularly on coding, including agentic coding and tool use tasks.
The company conducted joint pre-deployment testing of the latest Sonnet model with the US AI Safety Institute (US AISI) and the U.K. Safety Institute (UK AISI).
Anthropic also evaluated the model for catastrophic risks and reports it meets the ASL-2 Standard.
The upgraded Claude 3.5 Sonnet is available to all users.
Claude 3.5 Haiku
Claude 3.5 Haiku is the next generation of the company's lightest and fastest model.
The company reports Claude 3.5 Haiku improves performance across various skill sets and is particularly strong on coding.
Claude 3.5 Haiku will be released later in October.