Claude Chat

The Generic endpoint connects directly to standard pinned model infrastructure for predictable latency and SLA-backed deployments, whereas the Openrouter variant provides flexible routing across multiple upstream providers with instant fallback access during high traffic.

Choose the Generic endpoint for production applications requiring consistent low latency and pinned model versions like Claude Sonnet 4.6. Choose the Openrouter variant when prototyping, running cross-provider benchmarks, or needing dynamic fallback routing.

Select Claude Haiku 4.5 for high-speed edge tasks and lightweight sub-agents, Claude Sonnet 4.6 for the optimal balance of near-Opus reasoning and rapid execution, and Claude Opus (4.5, 4.6, or 4.7) when maximum logical density and strict instruction adherence are required.

Structure your inputs using XML tags such as <context>, <instructions>, and <example>, placing your final formatting instructions at the very end of the prompt. For deterministic tasks like coding or JSON generation, keep the temperature parameter near 0.0.

Claude Chat excels at complex code generation, autonomous agentic planning, structured JSON parsing, and 1-million-token document analysis. It is less suited for unfiltered edgy narratives or micro-tasks where ultra-low latency is the sole priority.