DeepSeek V4 Chat Model icon

DeepSeek V4 Chat Model

DeepSeek V4 como Chat Model para o AI Agent. Thinking Mode desativado por padrão.

DeepSeek V4 Chat Model

AI-generated

Summary

Provides a DeepSeek V4 chat model for connecting to an n8n AI Agent. Configure the model and optional thinking and sampling settings; when thinking is enabled, the node preserves reasoning data needed for tool-call follow-up requests.

Inputs

  • DeepSeek API credentials (required) — API key and optional base URL; defaults to https://api.deepseek.com.
  • Model — Choose DeepSeek V4 Flash (default) or DeepSeek V4 Pro.
  • Thinking Mode — Disabled by default; can be enabled with low, high, or max effort.
  • Return Thinking Content — Shown when thinking is enabled. If true, includes reasoning in generated text between tags; otherwise reasoning is omitted from visible text.
  • Max Tokens — Maximum response tokens; defaults to 2048.
  • Temperature — Available when thinking is disabled; defaults to 1 and is ignored when thinking is enabled.
  • Top P — Sampling diversity setting; defaults to 1. Applied only when thinking is disabled.
  • Frequency Penalty — Repetition penalty; defaults to 0. Applied only when thinking is disabled.
  • Presence Penalty — Penalty for tokens already present; defaults to 0. Applied only when thinking is disabled.

Output shape

A single AI language model connection for an AI Agent.

This operation provides a model connection, not generated text directly. Sampling settings are omitted from API requests when thinking is enabled.

Links

Discussion