Starta Anthropic, which develops artificial intelligence models competing with GPT-4 from OpenAI, has released Claude 3 Haiku. This is a new neural network in the Claude 3 family, according to the creators, which is three times faster than similar products in most workloads.
According to Anthropic, the speed of work makes Claude 3 Haiku ideal for applications where results are needed with minimal delays – for example, in customer support and answering questions. Haiku processes up to 21 thousand tokens (30 pages of text) per second with a request length of up to 32 thousand tokens.
“Corporate users pay special attention to speed – this is what helps them quickly analyze huge amounts of data and serve customers in a timely manner. The speed of Claude 3 Haiku allows you to quickly answer questions in chat and perform many small tasks at the same time,” the company said in a statement.
Anthropic has established a fairly loyal pricing policy, allowing large companies to save on processing huge amounts of data. The developers claim that Claude 3 Haiku will easily analyze 400 US Supreme Court cases or 2,500 images for just $1.
“Haiku is able to analyze huge volumes of documents, for example, quarterly reports, contracts, court cases – at half the price and at the same speed as competitors,” emphasizes Anthropic PBC.
A family of three large language models, Claude 3, was introduced in March. According to the developers, the most advanced of them, Claude 3 Opus, has computing power comparable to the advanced developments of industry leaders such as OpenAI and Google.
If you notice an error, select it with the mouse and press CTRL+ENTER.