Amazon has announced that the new large language model (LLM) from Anthropic, Claude 3.5 Sonnet, is now available today in Amazon Bedrock, a service which lets developers deploy generative AI applications quickly. According to Neowin’s coverage of Claude 3.5 Sonnet, this model surpasses even the high-end Opus model of Claude 3.
Benchmarking data shows that Claude 3.5 Sonnet scored better on various benchmarks compared to Claude 3 Opus, GPT-4o, Gemini 1.5 Pro, and an early snapshot of Llama-400B. The only areas where it scored worse were solving maths problems and in a specific version of the undergraduate-level knowledge benchmark, MMLU.
Amazon said that DoorDash, WPP, and Swindon Borough Council in the UK were already using Claude 3 models on Amazon Bedrock to deliver their services. DoorDash has utilised generative AI to build a contact centre for customer support; WPP, the world’s largest ad company, is using it in its intelligent marketing operating system, WPP Open; and Swindon Borough Council is using it to convert complex documents into a more accessible format for people with learning disabilities.
With Claude 3.5 Sonnet, developers should see an improvement in written content which understands nuance and humour, sounds more natural, and has a human-like tone that sounds more authentic and relatable. It can also process images to interpret charts and graphs, has improved coding capabilities, understands context better, and can help with data science and analysis.
Despite Claude 3.5 Sonnet scoring better than Opus on many benchmarks, it costs a fifth of the price to run. Anthropic no doubt has Haiku and Opus variants of Claude 3.5 ready to drop soon and the Opus model will likely get back the crown in terms of its abilities.
Let us know in the comments if you"ve had time to give Claude 3.5 Sonnet a go yet. Have you noticed any substantial improvements over the previous version?
Source: Amazon