Claude 3.5 Upgraded: Coding Champs and 'Computer Use' Beta

Today, Anthropic supercharged Claude 3.5 Sonnet, boosting its SWE-bench score from 33.4% to 49.0%, leaving OpenAI in the dust. Enter Claude 3.5 Haiku, matching the former largest model Claude 3 Opus at similar costs. The quirky new 'computer use' feature is now in public beta, letting Claude click and type like a human—experimental but already outperforming rivals on OSWorld. Available via Anthropic API, Amazon Bedrock, and Google Cloud, these upgrades promise smarter, safer AI antics.