Skip to main content

Model Management

omx supports multiple LLM providers with native protocol support for each.

Supported Protocols

ProtocolDescription
AnthropicMessages API with prompt caching, extended thinking, and streaming
OpenAIChat Completions API, compatible with any OpenAI-format endpoint
GeminiNative generateContent API with streaming and function calling
Responses APIOpenAI's Responses API with built-in tool orchestration

Quick Add with Providers

Add all models from a provider in one command:

omx --list-providers
omx --add-provider deepseek --api-key sk-...
omx --remove-provider deepseek

Built-in providers:

ProviderModels
zenmuxZenmux model list (Anthropic, Responses, Gemini, OpenAI formats)
deepseekDeepSeek Chat, DeepSeek Reasoner
moonshotKimi for Coding: Kimi-2.5 (Anthropic and OpenAI formats)
openrouterOpenRouter model list (OpenAI and Anthropic formats)
zhipuGLM-5.1, GLM-5, GLM-5-Turbo
minimaxMiniMax-M2.7, MiniMax-M2.7-Highspeed

Adding a Model Manually

You can also add models through the menu:

  1. Press Escape to open the menu
  2. Select Manage your model list
  3. Select Add a new model

Required Fields

FieldDescription
NameModel identifier sent to the API (e.g. gpt-4o, claude-sonnet-4-20250514)
NicknameDisplay name in omx UI
Provideropenai, anthropic, gemini, or responses
API KeyYour API key
API URLBase URL for the API
Context SizeMaximum context window in thousands of tokens

Example Configurations

Anthropic

Name: claude-sonnet-4-20250514
Nickname: Claude Sonnet 4
Provider: anthropic
API Key: sk-ant-...
API URL: https://api.anthropic.com
Context Size: 200

OpenAI

Name: gpt-4o
Nickname: GPT-4o
Provider: openai
API Key: sk-...
API URL: https://api.openai.com/v1
Context Size: 128

Google Gemini

Name: gemini-2.5-flash
Nickname: Gemini 2.5 Flash
Provider: gemini
API Key: AIza...
API URL: https://generativelanguage.googleapis.com/v1beta
Context Size: 128

Kimi for Coding

Name: kimi-for-coding
Nickname: Kimi for Coding: Kimi-2.5
Provider: anthropic
API Key: your-kimi-key
API URL: https://api.kimi.com/coding/v1/messages
Context Size: 200

Ollama (Local)

Name: qwen2.5-coder:32b
Nickname: Qwen Coder
Provider: openai
API Key: ollama
API URL: http://localhost:11434/v1
Context Size: 32

Default Model and Agent Model

Default Model - The model used when starting a new session. Set via the configuration menu.

Agent Model - A secondary model used for agent tools (Agent_Explore, Agent_Glance, Agent_Search), web search, and git commit generation. Using a faster, cheaper model here keeps costs down while your main model focuses on reasoning.

If no agent model is set, the default model handles everything.

Switching Models

Use /model during a session to switch models. Switching between different API types resets the session.