Claude 3: Anthropic presents Claude 3, the new AI that aims to overcome Gemini and ChatGPT

Redefining AI Standards: Introducing Claude 3 by Anthropic

Published on
3 read
Claude 3: Anthropic presents Claude 3, the new AI that aims to overcome Gemini and ChatGPT

Claude 3 - Discover the revolutionary AI assistant, Claude 3, enhancing productivity& efficiency. Claude 3 lets anyone experience seamless integration of cutting-edge AI technology. Unleash the power of Claude 3 for optimized workflows & unparalleled results.

Anthropic, the company of artificial intelligence founded by former OpenAI employees announces the arrival of Claude 3, a family of AI models that, according to the company, equals or exceeds the performance of the top models currently in vogue. Unlike previous versions, Claude 3 AI is also "multimodal", that is, he can understand both texts and images.

Anthropic argues that Claude 3 AI will be able to answer multiple questions, understand longer instructions, and be more precise. The model is also able to process more information thanks to its ability to better understand the context. The Claude 3 AI family is made up of three models: Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus, with Opus representing the largest and "smarter" model".

The company claims that Opus and Sonnet are already available on claude.ai and through its API while Haiku will be released soon. All three models can be used for chatbots, automatic completion, and data extraction activities.

Previous versions of Claude AI refused to respond to some harmless prompts, which, according to the company, "suggest a lack of understanding of the context". The new models are less likely to refuse these types of requests.

Anthropic states that Claude 3 AI models can provide results almost instantly, even when analyzing complex material, such as research articles. The corporation’s site post describes Haiku, the parallel version of Claude 3 AI, as "the quickest and most affordable model on the market", and can read a search article filled with charts and tables "in less than three seconds".

Anthropic also claims that Opus has passed most of the models in numerous benchmarking tests. The model demonstrated reasoning at a postgraduate level higher than that of OpenAI's GPT-4, obtaining 50.4% in this test compared to 35.7% of GPT-4. Claude 3 also showed better skills in solving mathematical problems, writing code, and understanding logical reasoning.

The new AI Prompt Engineer models also present significant improvements compared to the previous Claude 2.1 model. Sonnet, the mid-range model, is twice as fast as Claude 2 and Claude 2.1. "Excels in activities that require quick responses, such as information retrieval or sales automation", says Anthropic.

The Claude 3 models have been trained on a combination of non-public internal data sets, third-party data, and public data available in August 2023. The company states, in a document introducing the three models, that the training was carried out using Amazon Web Services hardware (AWS) and Google Cloud. Both companies invested in Anthropic, with Amazon paying $ 4 billion to the company. Claude 3 AI will be available on AWS's Bedrock model library and Google's Vertex AI.

Nick Wels
Nick Wels

Nick Wels 89 Lancaster Rd, London W11 1QQ Contact Email: info@promptengineer-1.xyx

Discussion (0)