Anthropic is rolling out its latest generative AI (GenAI) technology, Claude 3, asserting its superiority over OpenAI’s Chat GPT-4, as per a press release dated March 4.

The new suite of models, including Claude 3 Haiku, Claude 3 Sonnet, and the most advanced, Claude 3 Opus, is positioned to compete with OpenAI’s GPT-4 in terms of performance metrics.

Backed by substantial venture capital investments, Anthropic’s announcement emphasized Claude 3’s improved analytical and predictive capabilities, claiming superior performance on certain benchmarks compared to GPT-4 (excluding GPT-4 Turbo) and Google’s Gemini 1.0 Ultra (excluding Gemini 1.5 Pro).

An important advancement is Claude 3’s multimodal functionality, enabling it to process and analyze text and images, aligning it with features found in some GPT-4 and Gemini versions. This includes the ability to handle up to 20 images in a single inquiry, designed to facilitate comprehensive image analysis.

However, Anthropic has restricted Claude 3’s image processing to address potential ethical and legal concerns, explicitly avoiding individual identification. The company has acknowledged limitations in the model’s performance with low-quality images and tasks requiring spatial reasoning or precise object counting.

Claude 3 primarily focuses on image analysis rather than artwork creation. The model reportedly enhances its predecessors' ability to follow multi-step instructions, generate structured outputs in formats like JSON, and provide multilingual support with an improved understanding of user requests.

With initial support for a 200,000-token context window, Claude 3 aims to deliver more contextually rich responses, comparable to Google’s latest Gemini model, which also supports a significant token context window.

Anthropic has transparently addressed Claude 3's limitations, acknowledging its vulnerability to biases and hallucinations—common issues in current GenAI models. The model lacks web search capabilities, limiting its access to data before August 2023. While it offers multilingual support, certain languages pose challenges due to resource constraints.

The company plans regular updates to enhance Claude 3’s capabilities and address its limitations. Opus and Sonnet models are currently available, with the Haiku model slated for release later this year. Anthropic has also introduced a pricing structure for these models, aiming to accommodate a diverse user base.

In a broader context, Anthropic aims to lead in developing AI self-teaching algorithms, potentially revolutionizing virtual assistants and other AI applications. The company is also exploring “constitutional AI” to ensure AI behavior aligns more predictably with human intentions.