Bug Report: DeepSeek Chat Model Node Missing V4 Parameters

Feedback
1

Summary

The DeepSeek Chat Model node (@n8n/n8n-nodes-langchain.lmChatDeepSeek) is incompatible with DeepSeek’s V4 models (deepseek-chat and deepseek-reasoner) due to missing support for the new reasoning_effort parameter introduced in their latest API update.

Environment

  • n8n version: [Your version]

  • Node: @n8n/n8n-nodes-langchain.lmChatDeepSeek v1

  • DeepSeek Model: deepseek-chat (V4)

Issue Description

DeepSeek recently updated their API for V4 models, introducing a new parameter called reasoning_effort to control the model’s reasoning capabilities. This replaces the previous approach of having a separate deepseek-reasoner model.

The current n8n DeepSeek node does not expose this parameter, making it impossible to properly configure V4 models. The node has also removed options for older model versions, forcing users to use V4 models without the ability to configure them correctly.

Expected Behavior

The DeepSeek Chat Model node should include a reasoning_effort parameter option (similar to how temperature, maxTokens, etc. are exposed) to allow proper configuration of V4 models.

Actual Behavior

  • The reasoning_effort parameter is not available in the node’s options

  • Workflows that previously worked with older DeepSeek models now fail or behave incorrectly with V4 models

  • No workaround exists within the UI to manually add this parameter

Steps to Reproduce

  1. Create an AI Agent workflow using the DeepSeek Chat Model node

  2. Select deepseek-chat (V4 model)

  3. Attempt to configure reasoning capabilities

  4. Observe that reasoning_effort parameter is not available in the node options

Impact

  • Breaking change: Existing workflows using DeepSeek models are broken

  • No fallback: Older model versions have been removed from the node, leaving no working alternative

  • Blocking: Users cannot properly utilize DeepSeek V4 models in n8n

Proposed Solution

Add reasoning_effort parameter to the DeepSeek Chat Model node options with:

  • Type: String or Number (based on DeepSeek API spec)

  • Description: “Controls the reasoning depth for V4 models”

  • Optional: Should have a sensible default value

Additional Context

According to DeepSeek’s documentation:

“The V4 models have changed how their reasoning capabilities work. They have introduced a new parameter called reasoning_effort to control this, instead of simply having a separate deepseek-reasoner model”

Urgency

High - This is a breaking change that affects all users of DeepSeek models in n8n. The removal of older model options without proper V4 support leaves users with no working configuration.

1 Like
2

Commenting to receive updates