Anthropic reveals Claude 3 AI LLMs, claims it beats Gemini and GPT-4 in many benchmarks

Anthropic, the startup AI large language model (LLM) company that was launched in part by a number of former OpenAI team members, has just launched its latest family of LLMs. The Claude 3 lineup includes one that the company claims beats OpenAI's GPT-4 and Google's Gemini 1.0 Ultra in many AI benchmarking tests.

In its press release, Anthropic says the highest-end LLM, Claude 3 Opus, beats OpenAI and Google's models in benchmark questions like undergraduate-level knowledge, common knowledge, grade school math, and more. In addition to better chatbot responses, the company has added multimodel features to the Claude 3 models. It says users can upload images, documents, charts, and more to the Claude 3 models and they can handle responses to questions about that content that are "on par with other leading models."

Claude 3 Opus is the most advanced model, but Anthropic says that Claude 3 Sonnet offers solid performance at a lower cost, and is best used for enterprise operations. The smallest model is Claude 3 Haiku and is designed to offer fast responses to simpler questions. CNBC reports that Claude 3 can summarize up to 150,000 words, up from 75,000 words from the older Claude 2, which launched less than a year ago in June 2023.

Anthropic says both Claude 3 Opus and Sonnet are available right now in 159 countries., with Haiku coming in the near future. People can try out the Sonnet-based chat or free now via claude.ai. The Opus chatbot is available with a paid subscription to Claude Pro for the price of $20 a month.

Since launching in 2021, Anthropic has raised money for its AI LLM work from major tech companies including Amazon and even one of its big AI rivals Google. CNBC says it has raised a total of $7.3 billion from investors.

Source