Anthropic unveiled the Claude 3.5 Sonnet model, an upgrade from the recently launched Claude 3 model in March. The standout feature is the introduction of the mid-sized Sonnet model, surpassing the massive Claude 3 Opus model that was introduced earlier.
Testing results show that the Cluade 3.5 Sonnet model outperformed almost every test set, with only a slight defeat to MMLU against GPT-4o and the MATH test set against GPT-4o as well. Anthropic specified that in-house programming test sets favored the Claude 3.5 Sonnet over the Claude 3 Opus by a significant margin of 64% to 38%, despite being up to 5 times more cost-effective. This model is a multi-modal type like its counterparts, and has proven to outperform nearly all models except for GPT-4o, which only lost to the MMMU test set.
For the Claude chat service, the addition of Artifacts feature allows users to separate code, web, or document creation from the chat itself, facilitating easier conversation and collaboration.
The Cluade 3.5 Sonnet model is now available for free on Claude.ai website, while API usage is priced at $3 for input per million tokens and $15 for output per million tokens, similar to the previous Claude 3 Sonnet model.
TLDR: Anthropic introduces Claude 3.5 Sonnet, a high-performing model that outshines its predecessors in various test sets, offering improved features and cost-effective solutions for users.
Leave a Comment