March 6, 2024: Anthropic, a number one AI startup backed by significant investment from Google and enterprise capital, has announced the discharge of its latest GenAI technology. Claude3.
This recent family of models, consisting of Claude 3 Haiku, Claude 3 Sonnet and Claude 3 Opus – the latter being probably the most powerful – is designed to outperform OpenAI's GPT-4, particularly within the areas of study, prediction and specific benchmark tests against other major AI models.
Claude 3 marks Anthropic's entry into multimodal GenAI that may understand each text and pictures.
This enhancement allows Claude 3 to handle various document types, including photos, charts, graphs and technical diagrams, distinguishing it from some versions of GPT-4 and Google Gemini 1.0 Ultra.
However, for ethical and legal reasons, Anthropic has decided to limit certain features, similar to image generation and the identification of individuals in images.
The model is designed to deliver more expressive and interesting responses, higher follow multi-step instructions, and produce leads to structured formats.
It also features an enhanced context window that permits for a deeper understanding of inputs and generates more contextual responses.
Anthropic's latest AI, Claude 3, shows significant improvements in providing accurate answers, which is critical for firms serving customers across multiple industries.
By evaluating Claude 3's performance on a wide selection of complex, factual questions designed to uncover weaknesses in current AI models, Anthropic saw a major improvement in accuracy. Specifically, the Opus version of Claude 3 doubled the accuracy rate on these difficult questions in comparison with its predecessor, Claude 2.1, while reducing the variety of incorrect answers.
Here are the model prices:
| Model | Claude 3 Opus | Claude 3 Sonnet | Claude 3 Haiku |
|---|---|---|---|
| Cost | (Enter $/Million Token | Issue $/million tokens) $15 | $75 |
| Context window | 200,000* | 200,000 | 200,000 |
But Claude 3 also has its limitations. For example, it cannot search the Internet, which limits its answers to dates before August 2023, and it’s less fluent in certain languages than in English.
Anthropic acknowledges existing challenges similar to bias and hallucinations, but guarantees regular updates to enhance Claude 3's capabilities.
Anthropic's vision goes beyond the present version of Claude 3 and goals to develop a complicated self-learning AI algorithm.
This claim is consistent with efforts to develop AI systems that may perform a wide selection of tasks, from administrative support to creative endeavors.
