Anthropic has tested its latest AI model, Claude 3.7 Sonnet, using the classic Game Boy game Pokémon Red.

A blog post released on Feb. 24 described how the model used basic memory, screen pixel inputs, and command functionalities to simulate gameplay.

The company reported that Claude 3.7 Sonnet showed improved problem-solving abilities compared to its predecessor, Claude 3.0.

Anthropic said the model performed 35,000 actions to reach its third gym leader opponent – out of the usual eight.

While Claude 3.0 had difficulty leaving the starting area in the game, the updated model successfully defeated three gym leaders and earned their badges.

Recent Anthropic developments