Claude Opus 4.1

Famille de modèles de Anthropic · utilisé par 2 applications

Modèle d'IA

Claude Opus 4.1 is an incremental update to Claude Opus 4, released by Anthropic in August 2025. It represents refinements in reasoning quality, instruction-following, and overall performance while maintaining a similar safety and capability profile to its predecessor. The model is optimized for advanced reasoning, agentic tasks, and complex workflows, with a focus on improved alignment and reliability. Claude Opus 4.1 was trained on a proprietary mix of publicly available internet data (up to March 2025), private datasets, opted-in user data, and synthetic data. It underwent RLHF (Reinforcement Learning from Human Feedback) and Constitutional AI fine-tuning to align with principles like helpfulness, honesty, and harmlessness. The model is text-only and multilingual. Safety evaluations show strong alignment, with a 98.76% harmless response rate on violative requests and low over-refusal rates (0.08%) on benign prompts. It was deployed under AI Safety Level 3 (ASL-3) protections, consistent with Claude Opus 4, as a precautionary measure.