The anthropier used to measure the latest model of artificial intelligence

Anthropor Pokemon used to measure the latest model of artificial intelligence. Yes, really.

In a blog mail It was published on Monday, and Antarubor said it had tested its last model, Claude 3.7 SonataOn Boy Boy Classic Pokémon Red. The company prepared the form with basic memory, inserting the screen pixels, and job calls to press the buttons and navigate the screen, allowing it to run Pokémon continuously.

A unique feature of Claude 3.7 Sonnet is its ability to engage in “expanded thinking”. Like Openai’s O3-MINI and DeepSeek’s R1, CLAUDE 3.7 Sonnet can think “through difficult problems by applying more computing-and taking more time.

It apparently came to my hands in Red Pokemon.

Compared to a previous version of Claude, Claude 3.0 Sonit, who failed to leave the house in the town of Pleet where the story begins, Claude 3.7 Sonnet fought with three Pokemon Sports Pokemon leaders and won their badges.

Now, it is not clear the amount of computing required for Claude 3.7 Sonnet to reach those features – and the duration of each of them. The human said only that the model conducted 35,000 procedures to reach the last gym captain, which retreated.

Certainly, it will not take place long before some of the adventurous developers discovered.

Pokémon Red is more than a game more than anything else. However, there He is Long history One of the games used for measurement of artificial intelligence. In the past few months alone, a number of new applications and platforms have emerged to test games playing on the addresses that range between them Street fighter to pisionary.