Latest posts about Anthropic

About Anthropic

Anthropic is best known as the maker of the “human-centric” chat model, Claude. Anthropic was started by a group of OpenAI employees worried about the future of AI and its effects on human society. They were key members of the AI research community who had worked on AI safety and ethics.

If that mission sounds familiar, it’s because it is the same goal as OpenAI itself. But the founders of Anthropic weren’t satisfied with the direction OpenAI was taking. So they established Anthropic to focus on AI’s impact from a human perspective and to help guide the alignment of AI with human goals.

Claude, Claude 2, and Claude 3

The Claude models are known for being highly conversational. In general, their reasoning and coding abilities aren’t on par with GPT-4, but their language abilities are quite comparable.

Claude-100k was one of the first models released with a 100k+ context window, dwarfing the 4096-token window of the first GPT-3 model and others released thereafter, and even the impressive GPT-4-32k (which wasn’t yet released at the time).

Claude 3 was released in March 2024. It comes in three model sizes, name Haiku, Sonnet, and Opus. Opus is reported to slighttly outperform GPT-4 in some benchmarks, a major achievement with OpenAI having held the top spot since the beginning of the generative AI boom. The mid-size model, Sonnet, compares very favorably to GPT-3.5.

The ability to accurately summarize and work with such large amounts of text set Anthropic’s Claude models apart and has made them popular with writers.