Anthropic's Claude 3 is better and stronger than any other AI currently on the market

A little over half a year ago, Anthropic, an AI company, announced the release of the Claude 2 AI model, which at that time was the most powerful AI on the market at its release. Well, approximately 7 months later, Anthropic claims that it repeated the achievement with the release of its new AI model, Claude 3.

The company announced the release of Claude 3 in no less than 3 versions: Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku, each one with its specificities. When it comes to raw computing power, however, Claude 3 Opus is the strongest of them all, similar to GPT-4 Turbo, or Copilot Pro, for that matter.

Claude 3 is a multi-modal AI, meaning it is capable of working with text-based inputs, as well as visual-based inputs, as most current AI models do now.

According to Anthropic, Claude 3 Opus manages to outperform all the other existing AI models on the market, when it comes to common evaluation benchmarks for AI systems, including undergraduate-level expert knowledge (MMLU), graduate-level expert reasoning (GPQA), basic mathematics (GSM8K), and more, as you can see in the image below.

However, when it comes to visual understanding, Claude 3 is slightly behind Google’s Gemini 1.0 Ultra.

The model is nonetheless faster and more accurate than its previous model and performs tasks at a much faster time, overall.

Anthropic has chosen to dissect Claude 3 into 3 different versions, possibly following the competition’s model: there are already 2 versions of the Copilot and 3 versions of the Gemini AI.

Claude 3 Opus is the most capable model of the 3, and Anthropic describes it as best-in-market performance on highly complex tasks. It can also navigate open-ended prompts and sight-unseen scenarios with remarkable fluency and human-like understanding. It has a context length of 200K and it costs $15 | $75 (input/output).

Claude 3 Sonnet is described as the balance between intelligence and speed, and it is aimed at those customers looking to use it for scaling products and services. It has a context length of 200K and it costs $3 | $15 (input/output).

Finally, Claude 3 Haiku is aimed at personal users, as it is capable of completing tasks at a faster speed than the other two versions, as long as the tasks are not complex. It has a context length of 200K and it costs $0.25 | $1.25 (input/output).

The Opus and Sonnet versions are now available to use immediately upon registering, but the Haiku version will be available at a later date.

You can read the full blog post here.