Hugging Face models

Choose the model behind your connection.

Keep the same RateLimitless endpoint in every coding tool. Pick the model that fits your work, then change it later without touching your tool settings.

Front-end preview

One click saves your preference in this browser only. No model weights are downloaded, no inference worker starts, and no GPU spend is created.

Curated coding models

Six public Hugging Face models, from fast edits to advanced agent workflows.

Loading saved choice...

  • Fast edits

    Qwen2.5 Coder 1.5B

    Qwen/Qwen2.5-Coder-1.5B-Instruct

    A small code-focused model for quick changes, autocomplete, and lightweight local setups.

    Qwen1.5BApache 2.0Model page
  • Compact reasoning

    Phi-4 Mini

    microsoft/Phi-4-mini-instruct

    A compact general model with strong reasoning for Python and everyday coding tasks.

    Microsoft3.8BMITModel page
  • Long context

    Granite 8B Code 128K

    ibm-granite/granite-8b-code-instruct-128k

    A coding assistant built for understanding larger files and longer repository context.

    IBM8BApache 2.0Model page
  • Code + math

    DeepSeek Coder V2 Lite

    deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct

    A mixture-of-experts coding model for deeper implementation and technical reasoning.

    DeepSeek16B MoEDeepSeek licenseModel page
  • Agentic work

    gpt-oss-20b

    openai/gpt-oss-20b

    An open-weight reasoning model for tool use, multi-step work, and general coding agents.

    OpenAI21B MoEApache 2.0Model page
  • Advanced agentsFeatured

    Qwen3 Coder 30B

    Qwen/Qwen3-Coder-30B-A3B-Instruct

    A larger agentic coding model for repository-scale changes and tool-driven workflows.

    Qwen30B MoEApache 2.0Model page