Skip to content

Anthropic unveils Claude 3 family surpassing ChatGPT and Gemini

  • by
  • 3 min read

Anthropic, one of the leading players in the artificial intelligence (AI) landscape, has unveiled the Claude 3 model family. The new series, comprising the Haiku, Sonnet, and Opus models, offers much more intelligence and speed, surpassing GPT 4 and Gemini.

Opus and Sonnet are available in claude.ai and the Claude API in 159 countries. The Haiku model is expected to join the lineup soon. Sonnet is also available through Amazon Bedrock and in private preview on Google Cloud’s Vertex AI Model Garden.

The Claude 3 family initially offers a 200K context window, but the models can handle inputs exceeding 1 million tokens. Here, Opus takes the lead with a near-perfect recall, surpassing 99% in the ‘Needle In A Haystack’ evaluation.

Opus model in Needle In A Haystack evaluation. | Source: Anthropic

One of the standout features of the Claude 3 models is their near-instantaneous response capability. These models are suitable for situations where immediacy is paramount, such as live customer tasks and several real-time interaction scenarios.

The models are targeted for specific customers and workflow. Opus, the flagship model, emerges as the most intelligent in the Claude 3 family. It outperforms its predecessors and peers on various evaluation benchmarks for AI systems showcasing near-human levels of comprehension and fluency across complex tasks.

Claude 3 family offers more intelligence capability than other AI models. | Source: Anthropic

Opus matches the speed of Claude 2 and 2.1 but with enhanced intelligence.

The Haiku model, claimed by the company based on various tests, is the fastest and most cost-effective in its intelligence category and can process information-dense research papers in a record time.

Sonnet, on the other hand, is twice as fast as its predecessors and excels at tasks requiring rapid responses, such as knowledge retrieval and sales automation.

Claude 3 family offers much more visual prowess than GPT4 and Gemini. | Source: Anthropic

Claude 3 models have sophisticated vision capabilities and can process diverse visual formats, including photos, charts, graphs, and technical diagrams. This feature can be quite helpful for enterprise customers with knowledge bases encoded in various formats like PDFs, flowcharts, or presentation slides.

Moreover, the Claude 3 models demonstrate a more nuanced understanding of requests and are less likely to refuse answer prompts that approach system guardrails. Regarding accuracy, Opus stands out with a twofold improvement in correct answers compared to Claude 2.1.

The models exhibit fewer biases than previous iterations and are designed to be user-friendly.

With generative AIs gaining traction, they are also showcasing various biases, as happened with Gemini AI. With the newer iterations, companies are researching ways to reduce biases while offering higher intelligence and speed.

In the News: Apple launches two MacBook Air models powered by M3 chip

Kumar Hemant

Kumar Hemant

Deputy Editor at Candid.Technology. Hemant writes at the intersection of tech and culture and has a keen interest in science, social issues and international relations. You can contact him here: kumarhemant@pm.me

>