Anthropic launched Claude 4 Opus and Claude 4 Sonnet on Thursday. The company made bigger models after months of working on smaller ones. Alex Albert from Anthropic said companies want AI that can work alone on tasks. These models can code and solve problems for hours without stopping. Earlier versions only worked for one or two hours before making mistakes.
The models remember important information during long work sessions. They create files to track their progress, just like people take notes. GitHub picked Sonnet 4 for its coding helper instead of choosing a Microsoft model. Companies testing the models report better results than before. Opus 4 scores highest on coding tests, and Sonnet 4 works faster.
Both models cost the same as older versions. Opus charges fifteen dollars for every million words it reads and seventy-five dollars for what it writes. Free users can try Sonnet 4 but need to pay for Opus 4. People can use these models through Anthropic's website and other services. The company fixed most problems where models did things users didn't ask for.
The models remember important information during long work sessions. They create files to track their progress, just like people take notes. GitHub picked Sonnet 4 for its coding helper instead of choosing a Microsoft model. Companies testing the models report better results than before. Opus 4 scores highest on coding tests, and Sonnet 4 works faster.
Both models cost the same as older versions. Opus charges fifteen dollars for every million words it reads and seventy-five dollars for what it writes. Free users can try Sonnet 4 but need to pay for Opus 4. People can use these models through Anthropic's website and other services. The company fixed most problems where models did things users didn't ask for.
