Anthropic has introduced Claude 4, the latest version of its large language model, and says it performs better on difficult reasoning tasks than earlier releases. The company also says the new model includes additional safety measures designed to reduce misuse.

The launch comes as major AI developers continue to compete on performance, reliability, and guardrails. Anthropic is positioning Claude 4 as a step forward not only in capability, but also in attempts to make the system less likely to be used for harmful purposes.

According to the company, the model’s upgrades focus on handling more complex prompts and improving behavior in sensitive settings. As with other recent AI releases, the real-world impact will depend on how the model performs outside company testing and how customers put it to use.