While the race for productive artificial intelligence continues unabated, we are seeing new steps in this area from many companies. Now it was founded by former employees of OpenAI, the company behind ChatGPT, in which Google has invested billions of dollars. Anthropicpresented us with its latest models.

“New Models”claude 3” is given the name. The Claude 3 family, consisting of 3 versions, offers improvements in many areas over previous generations of Claude. Anthropic even claims that the models are better than Google’s Gemini and OpenAI’s ChatGPT.

Can answer more questions and understand longer instructions

Claude 3, the language model family, unlike its predecessors a multiple model It will appear as . So just not only text but also images will be able to understand. According to Anthropic’s statement, in this family Claude 3 Haiku, Claude 3 Sonnet and Claude 3 Opus Three models are available:

Opus will be the most advanced of these. It is made available using the Sonnet and Opus APIs. While Sonnet offers a free experience, Opus is offered exclusively to developers for Claude Pro members. Haiku will debut later. All three models chatbots, data extraction, autocomplete It can be used in vehicles and tasks such as.

Anthropic’s Claude models are still not available in Turkey. Therefore, we cannot access Claude 3 from our country. Although the company says it is working to bring it to other countries, there is no exact information about when it will arrive.

Anthropic, Claude 3 can answer more questions, understand longer instructions, and provide more accurate answers say. There is also an increase in the context window. So it will be able to process more information. According to the information received, first of all, all three members of the family were killed. Context window of 200,000 tokens will present. However, this number could be as high as 1 million. Google Gemini’s context window, which was introduced in mid-February, could also reach 1 million. This meant processing 700,000 words at once.

The Haiku model, which will be released at a later date, is announced by the company as “The fastest and most cost-effective model on the market” It was described as. The description explains that the model contains a large research paper full of charts and graphs. Can read in less than 3 seconds It was claimed.

Previous Claude models could refuse to answer questions that were considered ‘harmless’. The company stated that this has been resolved with the new generation and that the model is much less likely to reject commands. It was also reported that he could answer difficult questions correctly much faster than Claude 2.1.

Tests showed it was better than GPT-4

Anthropic also published comparisons of Claude 3 with GPT-4 and Gemini 1.0, which powers ChatGPT. In the data, Claude 3 reasoning It was seen to be better than GPT-4 (35.7%) by 50.4%. on the other hand math, coding, general knowledge, problem solving It was reported that Claude 3 gave better results than Gemini and GPT-4 in many categories such as.