Skip to content

Supported Providers and Models

💡 Quick Tip: Use the convenient model@provider syntax to specify both model and provider in one argument: janito -m gpt-4@openai "Your prompt"

This page lists the supported providers and their available models, organized by openness and sorted alphabetically within each category.

📋 Platform & Documentation Access: For information about which providers have publicly accessible platforms and documentation vs. blocked access, see Platform & Documentation Access.

Open-Source / Open-Weight Models

Alibaba

  • Qwen3 235B A22B Instruct (default) - Latest 1M context model
  • Qwen3 235B A22B Thinking - Reasoning-focused version
  • Qwen3 30B A3B Instruct - Compact 1M context model
  • Qwen3 30B A3B Thinking - Compact reasoning version
  • Qwen3 Coder Plus - Specialized for programming tasks
  • Qwen3 Coder 480B - Large-scale coding model
  • Qwen Turbo - High-speed general purpose
  • Qwen Plus - Balanced performance
  • Qwen Max - Maximum capability
  • Qwen3-Max-Thinking (2026-01-23) - Latest flagship reasoning model with adaptive tool use and test-time scaling

DeepSeek

  • DeepSeek Chat (default) - General purpose chat model (128K context)
  • DeepSeek Reasoner - Specialized for complex reasoning tasks (128K context)

MiniMax

  • MiniMax M2.1 (default) - Latest model with 230B total parameters (10B activated per inference), optimized for code generation and refactoring with polyglot code mastery and enhanced reasoning (128K context)
  • MiniMax M2.1 Lightning - Same performance as M2.1 with significantly faster inference and low latency (128K context)
  • MiniMax M2 - Previous generation model with 200k context length, 128k max output, agentic capabilities, function calling, and advanced reasoning

MiniMax models are optimized for coding, tool use, instruction following, and long-horizon planning. M2.1 excels in multilingual scenarios and matches or exceeds Claude Sonnet 4.5 performance.

Mistral

  • Mistral Large Latest (default) - Most capable Mistral model with 128k context
  • Mistral Medium Latest - Balanced performance with 32k context
  • Mistral Small Latest - Compact and efficient with 32k context
  • Codestral Latest - Specialized for code generation with 256k context
  • Codestral 2405 - Previous version of code-focused model
  • Devstral Small Latest - Optimized for agentic tool use in software development
  • Devstral Medium Latest - Enhanced agentic capabilities for development tasks

Mistral provides both general-purpose and specialized models, with Codestral specifically designed for code generation and Devstral for agentic software development.

For setup instructions, see the Mistral Setup Guide.

Moonshot

  • Kimi K2 0905 Preview (default) - Latest generation with enhanced performance
  • Kimi K2 Turbo Preview - Turbo version with optimized speed
  • Kimi K2 0711 Preview - Previous preview version

Moonshot provides open-source Kimi models with competitive performance.

Z.AI

  • GLM-4.5 (default) - Advanced reasoning and conversation
  • GLM-4.5 Air - Compact and efficient version

IBM WatsonX

Open-Source Models:

  • openai/gpt-oss-120b (default) - Open-source 120B model with thinking capabilities
  • openai/gpt-oss-20b - Open-source 20B model with thinking capabilities

IBM Granite Models:

  • ibm/granite-3-8b-instruct - IBM's Granite 3 8B Instruct model with 128K context
  • ibm/granite-3-3-8b-instruct - Updated Granite 3.3 8B Instruct model

Hosted Models:

  • meta-llama/llama-3-1-70b-instruct - Meta Llama 3.1 70B hosted on WatsonX
  • meta-llama/llama-3-3-70b-instruct - Meta Llama 3.3 70B hosted on WatsonX
  • mistralai/mistral-large - Mistral Large model hosted on WatsonX
  • mistralai/mistral-large-2407 - Mistral Large 2407 version

IBM WatsonX provides access to IBM's Granite models as well as popular open-source models hosted on their platform. All models support 128K context windows.

Proprietary Models

Anthropic

  • Claude Sonnet 4.5 (default) - Latest model with enhanced agent and coding capabilities
  • Claude Opus 4.1 - Exceptional model for specialized complex tasks
  • Claude Opus 4 - High-capability model for complex tasks
  • Claude Sonnet 4 - Balanced performance and capability
  • Claude 3.7 Sonnet - High-performance model with extended thinking
  • Claude 3.5 Haiku - Fast and efficient
  • Claude 3 Haiku - Compact and cost-effective

Google

  • Gemini 2.5 Flash (default) - Fast and efficient
  • Gemini 2.5 Pro - Advanced capabilities with extended context
  • Gemini 2.5 Flash-Lite Preview - Lightweight preview version

OpenAI

  • GPT-5 (default) - Latest generation with advanced reasoning
  • GPT-5 Mini - Compact version of GPT-5
  • GPT-5 Nano - Ultra-compact version
  • GPT-4.1 - Enhanced GPT-4 with improved capabilities
  • GPT-4.1-mini - Balanced performance and efficiency
  • GPT-4.1-nano - Lightweight version
  • GPT-4 Turbo - High-performance GPT-4 variant
  • GPT-4o - Multimodal GPT-4 optimized for chat
  • GPT-4o-mini - Compact multimodal version
  • o3 - Advanced reasoning model
  • o3-mini - Compact reasoning model
  • o4-mini - Latest mini reasoning model
  • GPT-3.5 Turbo - Cost-effective general purpose

For setup instructions, see the OpenAI Setup Guide.

Azure OpenAI

  • Azure OpenAI Deployment - Custom Azure OpenAI deployments

For setup instructions, see Using Azure OpenAI with Janito.