Create Model Profile

A Model Profile defines how a Large Language Model (LLM) behaves within Calibo Accelerate. It allows you to configure model selection, capabilities, response behavior, safety controls, cost settings, and token limits in a reusable and standardized way.

After you create a model profile, you can:

Set it as the default model profile for a product. This profile is applied automatically to tasks that do not have a specific profile assigned.
Assign it to task-specific use cases such as Agentic Workflow, Chat, Classification, Reasoning, and Summarization. In these cases, the assigned profile overrides the default profile.

Using model profiles helps you:

Ensure consistent AI behavior across products and workflows
Optimize model configurations for specific tasks and use cases
Maintain governance, safety, and cost control through centralized configuration

Model profiles provide flexibility to tailor AI behavior for different scenarios while maintaining control and standardization across your organization.

Prerequisites

Before creating a model profile, ensure that:

An LLM provider (such as OpenAI or Gemini) is already configured and activated.
The connection has been successfully tested.
Models are visible in the Models section after connection validation.
You have the required permissions to configure LLM settings in Calibo Accelerate.

Create a Model Profile

Follow these steps to create a model profile:

Sign in to the Calibo Accelerate platform and click Configuration in the left navigation pane.
On the Platform Setup screen, under the Cloud Platform, Tools & Technologies tile, click Configure.
On the Cloud Platform, Tools & Technologies screen, navigate to the Large Language Models section and click Configure.
(After you save connection details for at least one LLM provider, you will see the Modify button here.)
Locate the LLM provider connection details where you want to create the model profile (for example, OpenAI or Gemini).
Click Create Model Profile for the selected LLM.

On the Create Model Profile screen, enter the following details such as profile name, model, capabilities, and configuration settings.

Field	Description	Recommended Values / Use Cases
Profile Name	Enter a unique name to identify the model profile.	Use meaningful names like Customer Support Bot, Content Generator, Code Assistant.
Description	Enter a brief description of the model profile.	Describe purpose, e.g., Used for summarization and FAQs.
Model	Select the model to be used for this profile. Available models depend on the configured provider and account permissions.	Use advanced models (for example, GPT, Gemini Pro, or Claude Opus) for complex reasoning tasks; use lighter models for cost-sensitive workloads.
Capability	Select one or more capabilities supported by the selected model. Available capabilities vary by provider and model. Examples include Batch Processing, Citations, Code Execution, Context Management, Structured Outputs, Text Generation, Embeddings, and multimodal capabilities.	Select capabilities based on the intended use case. For example, use Code Execution for programming workflows, Citations for grounded responses, Structured Outputs for integrations, and Embeddings for semantic search.
Effort	Specifies the reasoning effort level used by supported models. Available only for providers and models that support reasoning controls.	Low: Faster responses. Medium: Balanced performance. High: Advanced reasoning, planning, analytics, and code generation.
Temperature	Controls randomness of the response. Lower values produce deterministic output; higher values increase creativity. Available only for models that support temperature settings.	0–0.3: deterministic (code, compliance) 0.4–0.7: balanced (chatbots) 0.8–1: creative (content writing)
Top-p	Controls diversity using nucleus sampling. Typically used as an alternative to temperature. Available only for supported models.	0.1–0.5: focused responses 0.6–1: diverse outputs (creative tasks)
Frequency Penalty	Reduces repetition of tokens in the response (range: -2 to 2). Available only for supported models.	0–1: reduce repetition (summaries) Negative: allow repetition (structured outputs)
Presence Penalty	Encourages introducing new topics in the response (range: -2 to 2). Available only for supported models.	0.5–1.5: brainstorming 0 or less: focused answers
Response Token Limit	Specifies the maximum number of tokens allowed in the response.	50–200: short replies 200–800: detailed answers 800+: long-form content
Response Format	Select the output format based on your use case such as Text, JSON Schema , JSON Object, or Tool.	Use JSON Schema/Object for structured APIs; Text for chat; Tool for integrations.
Safety Defaults	Enable Blocklist Categories: Turn on to restrict generation of content in sensitive categories. Select categories such as Hate, Sexual Content, Violence, Self Harm, and Politics.	Enable for enterprise/compliance use cases. Select based on compliance requirements (e.g., finance, healthcare). Select all for strict security, minimal for internal tools.
Safety Defaults	Enable Automatic Redaction: Select data types to redact, such as Email, Phone, Credit Card, and IP Address. Select how the data should be masked (for example, Mask, Replace or Hash).	Mask: general use Hash: audit/security Replace: readable logs
Costing	Pricing Source: Determines how token pricing is calculated for this profile, such as provider-defined Standard pricing or custom overrides. By default, Standard pricing is selected. If you select custom overrides , provide the following details: Input Token Rate: Specifies the cost per million input tokens for this profile. Used for cost estimation and reporting. Output Token Rate: Specifies the cost per million output tokens for this profile. Used for cost estimation and reporting. Currency: Select the currency (for example, USD) used for cost calculation and reporting.	Use Standard for simplicity; Custom Override for internal cost tracking.
Token Limits	Max Input Tokens: Specify the maximum number of tokens allowed in the input.	500–1,000 tokens: Chatbots, FAQs (short queries) 1,000–4,000 tokens: Document summarization (medium content) 4,000–10,000+ tokens: Large document processing (reports, logs, contracts)
Token Limits	Max Output Tokens: Specify the maximum number of tokens allowed in the output.	100–300 tokens: Short responses (chat replies, quick answers) 300–800 tokens: Detailed explanations (support, documentation) 800–2,000+ tokens: Long-form content (articles, analysis, reports)
Allowed Modalities	Select the modalities supported by the model that this profile can use (such as Text, Image, Audio, or Video).	Text: Chatbots, Q&A, documentation, code generation Image: Image analysis, captioning, visual search Audio: Speech-to-text, transcription, voice assistants Video : Video summarization, scene analysis
Input Modalities	Select the input types supported by the selected model. Available options vary by provider and model.	Text: Chatbots, Q&A, documentation. Image: Visual analysis and captioning. PDF: Document intelligence and summarization.
Output Modalities	Select the output types supported by the selected model. Available options vary by provider and model.	Text: Conversational AI and content generation. Structured Output: APIs, integrations, and workflow automation. Other modalities may be available depending on provider support.

Click Create to save the model profile.

The model profile is created and associated with the selected LLM configuration. It is now available for use across products and workflows.