Last night, AI startup Anthropic unveiled its newly upgraded Big Cup of Claude 3.5 Sonnet. Officially, improved coding capabilities are the biggest draw of Claude 3.5 Sonnet, which was tested by GitLab and found to have a 10% improvement in inference without additional latency, making it a great fit for multistep software development processes. the Browser Company also noted that Claude 3.5 Sonnet outperformed all previous models they tested in automating web workflows. The newly upgraded model, Claude 3.5 Haiku, which is the most responsive, also performed particularly well on coding tasks.
It scored 40.6% on SWE-bench Verified, outperforming many of the so-called state-of-the-art intelligences, including Claude 3.5 Sonnet (the older version) and GPT-4o. An upgraded version of Claude 3.5 Sonnet is now available to all users, and Claude 3.5 Haiku is due to be released this month. The upgraded Claude 3.5 Sonnet is now available to all users, while Claude 3.5 Haiku will be released later this month, initially supporting text input only, with image input to follow.
In addition, Anthropic has introduced a revolutionary feature called "computer use" for public testing. Simply put, through the API, Claude can now use standard tools and software designed by humans, and developers can use this to free up boring and repetitive process tasks, or even open-ended tasks such as research.