Models & Capabilities

AI models are the actual "brains" that work when you chat with AI. Different models have different capabilities and characteristics.

What is an AI Model

An AI model is a program trained on massive amounts of data that can understand and generate human language. Think of it as an assistant with vast knowledge, but each "assistant" has different specialties and abilities.

Core Capabilities of Models

Text Conversation

Basic capabilities that all models have, including:

• Answering questions
• Writing articles
• Translating text
• Writing code

Vision Capability

Some models support "seeing and describing images":

• Understanding image content
• Analyzing chart data
• Recognizing text (OCR)
• Describing image details

Models with vision: GPT 5.2, Claude 4.5 Sonnet, Gemini 3 Pro, etc.

Reasoning Capability

Deep thinking ability of advanced models:

• Solving complex math problems
• Multi-step logical reasoning
• Code debugging and optimization
• Scientific problem analysis

Enhanced reasoning models: o3, o4-mini (OpenAI), Claude 4.5 Sonnet, etc.

Tool Use

Enables AI to perform external operations:

• Searching web information
• Executing code
• Calling APIs
• Operating files

Note: Tool use requires application support. Chatbox currently supports some tool features.

Model Parameters Explained

When using models, you can adjust some parameters to control output behavior:

Temperature

Controls the "creativity" of responses:

Suggestion: Use the default value (around 0.7) for most cases.

Note: Many newer models (such as o-series reasoning models, some Claude models, etc.) do not support the Temperature parameter. If you encounter errors after setting it, set Temperature to "Not Set" in Chatbox settings.

Top P (Nucleus Sampling)

Another way to control randomness, usually used alternatively with Temperature:

• Top P = 1: Consider all possible words
• Top P = 0.1: Only consider the most likely 10% of words

Suggestion: Keep the default value unless you have specific needs.

Note: Similar to Temperature, many newer models also do not support the Top P parameter. If you encounter errors, set Top P to "Not Set".

Max Tokens (Maximum Output Length)

Limits the maximum length of AI responses:

• Smaller value (e.g., 500): Short answers
• Larger value (e.g., 4000): Detailed answers
• Not set: Let the model decide

Note: This is just an upper limit. AI will naturally end responses based on the question.

How to Choose Models

By Use Case

By Budget

Model Capability Quick Reference