Adding models
Open model dialogue
- Go to the model settings and click on Add Model to add a new model to the platform
- A modal opens where you can add models. Here, you find two sections:
- Display Settings at the top allows you to customize what the user sees in the model selector.
- Model Configuration lets you connect your Langdock workspace to your model API.
Display Settings
To configure the Display settings, you can follow the following steps. This information is also available by the company hosting the model. Provider: The organization that built and trained the model. This doesn’t necessarily align with the company you consume the model from. For example, you can use Microsoft Azure to use OpenAI models in the EU, but the provider will still be OpenAI. Model name: The name of the model. Hosting provider: Where you consume the model. For example, GPT-5 can be hosted by Microsoft Azure. Region: Shows the user where the model is hosted. This can be set to the EU, the US, or Global. Knowledge cutoff: When the model training data ended. Image analysis: Indicates whether the model can analyze images. This information is available from the model provider and the model hoster. Please only enable this setting if the model supports vision/image analysis. Model description: An optional description (max 150 characters) shown to users. You can provide descriptions in English and German using the language toggle.Model Configurations
To set up the Model Configuration, select the SDK you are using. You will find information on the configuration of the model provider (e.g., Azure or AWS): SDK: The kit or library Langdock needs to use the model you added. Base URL: To send prompts to the corresponding endpoint of your model. Model ID: The name of the model in your configuration (this might not be the “official” model name). API key: Allows your users to authenticate using the model from within Langdock when they send prompts. Context Size: The number of tokens the model can process in its context window. Please use the exact value of the model to ensure the context management in Langdock works correctly. API Type: Choose between Completion API and Responses API depending on the model. Newer models (GPT-5 series, o3, o4 Mini) typically use the Responses API, while older models use the Completion API.Other configuration options
Maximum messages in 3 hours: Allows you to influence usage/costs and limit messages per user. This setting is optional. Input and output token pricing: Allows you to set the token pricing of the individual model to monitor usage and costs. Reasoning Effort: Determines how much computation the model spends on reasoning. Higher values improve quality but incur extra latency and tokens. Accepted values: None, Minimal, Low, Medium, High. Only available for models using the Responses API. Verbosity: Controls the level of detail in the model’s final answer. Accepted values: Low, Medium, High. Only available for models using the Responses API. Canvas: Whether the model can generate canvas content for collaborative writing and editing. Supports Tools: Whether the model supports tool calling (integrations, web search, etc.). Enabled by default. Tool Calls in Stream: Whether the model can call tools during streaming. Enabled by default. Supports Temperature: Whether the model supports the temperature parameter for controlling randomness. Disable this for reasoning models (o1, o3, o4 Mini, GPT-5, GPT-5 mini, GPT-5 nano) as they reject temperature values. Claude models also don’t support temperature when Display Thinking is enabled. Display Thinking: Shows the model’s reasoning process in the UI. Enable this for reasoning/thinking model variants. Available in Agents: Whether this model can be selected for use in agents. Enabled by default. Visible to everyone: You can set the model to be visible to everyone in the workspace. If this option is disabled, the model is only visible to admins and cannot be used by other users. This allows you to test the model before launching it to the entire workspace. Maintenance mode: Can be activated to show users in the interface that the model might not work as expected. This is useful if you are changing some configuration or there is a temporary issue with the model from your model provider. Strict Mode: Enables strict validation mode for the model’s API calls. User Identifier: Controls whether user identity information is sent to the model provider. Options: None, User ID, Email, or Microsoft UPN (User Principal Name). Image Editing: For image generation models only — indicates whether the model can edit existing images in addition to generating new ones.Final steps
- After entering all mandatory settings, click Save
- We recommend testing the model before making it visible to everyone. Send a message to the model and see if there is a response generated by the model. If you run into any issues, contact support@langdock.com
Model-specific configuration
Below are the recommended settings for each model. Use these values when configuring models in Langdock.OpenAI GPT-5.2 models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| GPT-5.2 | Responses API | 400,000 | 128,000 | Set reasoning to minimal and verbosity to low |
| GPT-5.2 (Thinking) | Responses API | 400,000 | 128,000 | Set reasoning to high and verbosity to low |
| GPT-5.2 Pro | Responses API | 400,000 | 128,000 | Maximum reasoning depth and reliability |
OpenAI GPT-5.1 models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| GPT-5.1 | Responses API | 400,000 | 128,000 | Set reasoning to minimal and verbosity to low |
| GPT-5.1 (Thinking) | Responses API | 400,000 | 128,000 | Set reasoning to high and verbosity to low |
| GPT-5.1 Chat | Responses API | 128,000 | 16,384 | Only available at Azure as global standard deployment in EU or through OpenAI directly |
OpenAI GPT-5 models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| GPT-5 | Responses API | 400,000 | 128,000 | Set reasoning to minimal and verbosity to low |
| GPT-5 (Thinking) | Responses API | 400,000 | 128,000 | Set reasoning to high and verbosity to low |
| GPT-5 Chat | Responses API | 128,000 | 16,384 | — |
| GPT-5 mini | Responses API | 400,000 | 128,000 | — |
| GPT-5 nano | Responses API | 400,000 | 128,000 | — |
OpenAI GPT-4.1 models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| GPT-4.1 | Completion API | 1,047,576 | 32,768 | Good for data analyst (CSV/Excel files) |
| GPT-4.1 mini | Completion API | 1,047,576 | 32,768 | — |
| GPT-4.1 nano | Completion API | 1,047,576 | 32,768 | — |
OpenAI GPT-4o models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| GPT-4o | Completion API | 128,000 | 16,384 | — |
| GPT-4o Mini | Completion API | 128,000 | 16,384 | — |
OpenAI reasoning models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| o3 | Responses API | 200,000 | 100,000 | — |
| o3 Mini | Completion API | 200,000 | 100,000 | Use model ID o3-mini |
| o3 Mini high | Completion API | 200,000 | 100,000 | Use model ID o3-mini and set reasoning effort to high |
| o4 Mini | Responses API | 200,000 | 100,000 | — |
| o4 Mini high | Responses API | 200,000 | 100,000 | Use model ID o4-mini and set reasoning effort to high |
| o1 | Completion API | 200,000 | 100,000 | — |
Anthropic Claude models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| Claude Opus 4.6 | Completion API | 200,000 | 128,000 | — |
| Claude Opus 4.6 Reasoning | Completion API | 200,000 | 128,000 | Enable Display Thinking |
| Claude Opus 4.5 | Completion API | 200,000 | 64,000 | — |
| Claude Opus 4.5 Reasoning | Completion API | 200,000 | 64,000 | Enable Display Thinking |
| Claude Sonnet 4.6 | Completion API | 200,000 | 64,000 | — |
| Claude Sonnet 4.6 Reasoning | Completion API | 200,000 | 64,000 | Enable Display Thinking |
| Claude Sonnet 4.5 | Completion API | 200,000 | 64,000 | — |
| Claude Sonnet 4.5 Reasoning | Completion API | 200,000 | 64,000 | Enable Display Thinking |
| Claude Sonnet 4 | Completion API | 200,000 | 64,000 | — |
| Claude Sonnet 4 Reasoning | Completion API | 200,000 | 64,000 | Enable Display Thinking |
| Claude Sonnet 3.7 | Completion API | 200,000 | 8,192 | — |
| Claude Sonnet 3.7 Reasoning | Completion API | 200,000 | 8,192 | Enable Display Thinking |
| Claude Haiku 4.5 | Completion API | 200,000 | 64,000 | — |
Google Gemini models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| Gemini 2.5 Pro | Completion API | 1,000,000 | 64,000 | Supports reasoning/thinking |
| Gemini 2.5 Flash | Completion API | 1,000,000 | 32,000 | Supports reasoning/thinking |
| Gemini 3 Pro | Completion API | 1,000,000 | 64,000 | — |
| Gemini 3 Flash | Completion API | 200,000 | 32,000 | — |
Other models
| Model | API Type | Context Size | Max Output Tokens | Special Configuration |
|---|---|---|---|---|
| Mistral Large 2411 | Completion API | 128,000 | — | No image analysis support |
| Codestral | Completion API | 32,000 | — | Code generation model |
| Llama 4 Maverick | Completion API | 1,000,000 | — | — |
| Llama 3.3 70B | Completion API | 128,000 | — | No image analysis support |
| DeepSeek R1 | Completion API | 128,000 | — | US region only, no tool support |
| DeepSeek v3 | Completion API | 128,000 | — | — |
| Amazon Nova Lite | Completion API | 300,000 | — | US region only |
| Amazon Nova Pro | Completion API | 300,000 | 5,000 | US region only |
For the most up-to-date model information and capabilities, check the model picker in app.langdock.com.
Special cases during setup
Mistral from Azure: Make sure to select “Mistral” as the SDK. Claude from AWS Bedrock: When you select the Bedrock SDK, the “Base URL” field becomes “Access Key ID” and the “API Key” field becomes “Secret Access Key”. Enter your AWS credentials in these fields. See the Bedrock setup guide for details. Flux from Replicate: The base URL field needs to have the full model path, not just the base URL. For Flux 1.1 Pro this is:https://api.replicate.com/v1/models/black-forest-labs/flux-1.1-pro/predictions