Models

Models define which AI provider and configuration to use for prompts.

Basic Model Declaration

model myModel = {
  name: "claude-sonnet-4-20250514",
  apiKey: env("ANTHROPIC_API_KEY"),
  provider: "anthropic"
}

Configuration Fields

Required Fields

Field	Type	Description
name	text	Model identifier (e.g., “claude-sonnet-4-20250514”, “gpt-5.2”)
apiKey	text	API key for authentication
provider	text	Provider type: "anthropic", "openai", "google"

Optional Fields

Field	Type	Description
url	text	Custom API endpoint URL (defaults per provider)
maxRetriesOnError	number	Automatic retry count on API errors (default: 0)
thinkingLevel	text	Extended reasoning level for supported models
tools	array	Tools available to this model

Providers

Anthropic

model claude = {
  name: "claude-sonnet-4-20250514",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY")
}

OpenAI

model gpt = {
  name: "gpt-5.2",
  provider: "openai",
  apiKey: env("OPENAI_API_KEY")
}

Google

model gemini = {
  name: "gemini-3-flash",
  provider: "google",
  apiKey: env("GOOGLE_API_KEY")
}

Extended Thinking

Some models support extended reasoning with thinkingLevel:

model deepThinker = {
  name: "claude-sonnet-4-20250514",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY"),
  thinkingLevel: "high"
}

Levels: "none", "low", "medium", "high", "max"

Error Handling

Automatically retry on transient errors:

model resilient = {
  name: "gpt-5.2",
  provider: "openai",
  apiKey: env("OPENAI_API_KEY"),
  maxRetriesOnError: 3
}

Custom Endpoints

Use a custom API endpoint (OpenAI-compatible):

model local = {
  name: "llama-3",
  provider: "openai",
  apiKey: "not-needed",
  url: "http://localhost:11434/v1"
}

Models with Tools

Attach tools for the model to use:

import { readFile, writeFile, bash } from "system/tools"

model agent = {
  name: "claude-sonnet-4-20250514",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY"),
  tools: [readFile, writeFile, bash]
}

// Agent can now use these tools
vibe "Read the config file and update the version" agent

See Tools for more on tool definitions.

Multiple Models

Use different models for different tasks:

model fast = {
  name: "claude-haiku-4-5-20251001",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY")
}

model smart = {
  name: "claude-opus-4-5-20251101",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY")
}

// Quick tasks
let classification = do "Classify this: {text}" fast

// Complex reasoning
let analysis = do "Deeply analyze this: {text}" smart

Exporting Models

Share models across files:

config.vibe

export model claude = {
  name: "claude-sonnet-4-20250514",
  provider: "anthropic",
  apiKey: env("ANTHROPIC_API_KEY")
}

// main.vibe
import { claude } from "./config.vibe"
let result = do "Hello" claude