Anthropic releases Claude 4 with improved reasoning capabilities and new safety guardrails designed to reduce harmful outputs.
Anthropic has unveiled Claude 4, its most advanced and safest AI assistant yet, featuring significant improvements in reasoning, coding, and multimodal capabilities while maintaining the company's commitment to AI safety.
The new model demonstrates substantial improvements in complex reasoning tasks, with benchmark tests showing a 40% improvement over Claude 3 in mathematical problem-solving and a 35% improvement in coding tasks. Perhaps more importantly, the model shows significantly better calibration in acknowledging uncertainty.
"Claude 4 represents our commitment to building AI systems that are both capable and aligned with human values," said Anthropic's CEO. "We've made substantial advances in safety without sacrificing performance."
The release includes new safety features such as enhanced refusal capabilities for harmful requests, improved detection of potential jailbreak attempts, and a new "thinking" mode that allows users to see the model's reasoning process.
Anthropic has also introduced Claude Team, a new enterprise plan offering higher usage limits and dedicated support for business customers.
