Anthropic rivals OpenAI's 'magical' GPT-4o with impressive vision (and a great sense of humor)

 Anthropic Claude 3.5 Sonnet.
Anthropic unveils its new Claude 3.5 Sonnet model. | Credit: Anthropic

What you need to know

  • Anthropic unveiled Claude 3.5 Sonnet, available on the web or iOS.

  • The model is great at interpreting complex graphs and charts, translating code, and more.

  • It also features a new feature, Artifacts, which lets users see and interact directly with their results.

Microsoft, Google, and OpenAI are arguably at the forefront and most popular AI-powered chatbots. Even Apple ventured into the AI landscape recently with Apple Intelligence, which is set to give its date Siri AI assistant a much-needed overhaul.

Anthropic is among the key players in AI right now, and it's asserting its place in the landscape with a new AI model — Claude 3.5 Sonnet. The AI company compares its Sonnet model to OpenAI's GPT-4o or Google's Gemini and says it can perform at equal or even better standards. You won't have to wait for it to ship, as it's already available for web and iOS users. It's worth noting that Claude Pro and Team plan subscribers will have high rate limits compared to free users.

Claude 3.5 Sonnet boasts better reasoning and understanding capabilities. Interestingly, the AI model can read your handwriting, too. It also spots better performance than Claude 3 Opus per the benchmarks shared and maintains cost efficiency and high speeds (two times faster than its predecessor).

Anthropic benchmark results for Claude 3.5 Sonnet
Anthropic benchmark results for Claude 3.5 Sonnet

On paper, Claude 3.5 Sonnet is everything Anthropic promises it to be. It outperforms OpenAI's GPT-4o, Google's Gemini 1.5 Pro, and Meta's Llama 3 400B in 7/9 benchmarks. Perhaps more interestingly, it outperforms its rivals in 4/5 benchmarks accessing vision capabilities.

According to Anthropic:

"Sonnet now surpasses Claude 3 Opus across all standard vision benchmarks. Improvements are most noticeable in tasks requiring visual reasoning, like interpreting charts, graphs, or transcribing text from imperfect images."

This essentially places Anthropic's Claude 3.5 Sonnet model on an even or higher playing field with OpenAI's GPT-4o model. The benchmarks show the model's impressive capabilities in interpreting complex graphs and charts, translating code, and more. We already saw GPT-4o's coding capabilities at OpenAI's Spring Update event, which was impressive.

Additionally, Anthropic's new model is also good at transcribing text from images coupled with a great understanding of humor, which translates to better text-generated outputs.

More than just an AI-powered chatbot

Unlike the conventional way of interacting with AI-powered chatbots, Anthropic is doing things differently. Claude 3.5 Sonnet ships with a new feature dubbed Artifacts, which lets users see and interact with their results.

For instance, when using Microsoft Copilot to generate images, you'll need to transfer the image to another tool for edits and fine-tuning. While you can specify the edits and changes you want and ask the chatbot to make them, it doesn't always get the job done. However, Anthropic will now let users edit images in the app directly.

Anthropic plans to use the model in organizations to "securely centralize their knowledge, documents, and ongoing work in one shared space, with Claude serving as an on-demand teammate."