Claude 3 Opus

Claude 3 Opus is a large language model from Anthropic, the most capable variant of the Claude 3 family, introduced on 4 March 2024 alongside the Sonnet and Haiku models. It was positioned as Anthropic’s most intelligent model at launch, intended for complex tasks: research, strategy, analysis and automation.

The model is multimodal — it takes text and images (photos, charts, diagrams, documents/PDFs) as input and produces text responses. It supports a 200,000-token context window (with the ability to handle inputs exceeding 1 million tokens for select customers).

Claude 3 Opus was trained using reinforcement learning from human feedback (RLHF) and Anthropic’s Constitutional AI method. The model supports tool use, structured output and multilingual work.

The model was available commercially through the Claude API, the claude.ai platform, and Amazon Bedrock and Google Vertex AI. API pricing was USD 15 per 1M input tokens and USD 75 per 1M output tokens. Claude 3 Opus is now a legacy (deprecated) model, superseded by newer Claude models.