Claude (Anthropic)
model-providerClaude is Anthropic's family of large language models. The lineup includes Opus (most capable), Sonnet (balanced), and Haiku (fastest). Claude models are known for strong instruction following, 200K token context windows, and Constitutional AI safety training.
For agent systems, Claude offers tool use via tool_use content blocks, extended thinking mode for complex reasoning, and reliable structured output. The API is available directly from Anthropic or through aggregators like OpenRouter.
Claude excels at long-document analysis, careful reasoning, and tasks that require following complex multi-step instructions — all critical capabilities for autonomous agents.
Key Features
- •200K token context window
- •Tool use via content blocks
- •Extended thinking mode
- •Vision and document analysis
- •Constitutional AI safety
Pros
- +Excellent instruction following
- +Long effective context window
- +Strong at analysis and reasoning
- +Reliable tool use
Cons
- −Smaller ecosystem than OpenAI
- −No image generation
- −Fewer fine-tuning options