Anthropic has unveiled upgrades to its AI portfolio, introducing the enhanced Claude 3.5 Sonnet model, the new Claude 3.5 Haiku, and a "computer control" feature now in public beta.

The Claude 3.5 Sonnet has shown significant improvements across various benchmarks, with a standout performance in coding. It scored an impressive 49.0% on the SWE-bench Verified benchmark, outperforming all publicly available models, including those from OpenAI and other specialized coding systems.

In a groundbreaking move, Anthropic has also launched a computer control feature, enabling Claude to interact with computers like a human—viewing screens, controlling cursors, clicking, and typing. This functionality, now in public beta, makes Claude 3.5 Sonnet the first AI model to offer such capabilities.

Several leading tech companies have already begun integrating these features.

"Claude 3.5 Sonnet is a major advancement in AI-powered coding," said GitLab, reporting up to a 10% improvement in reasoning across use cases without any additional latency.

Set to launch later this month, the Claude 3.5 Haiku model matches the performance of Claude 3 Opus, while maintaining both speed and cost-efficiency. It achieved a 40.6% score on SWE-bench Verified, surpassing many competitors, including the original Claude 3.5 Sonnet and GPT-4o.