OpenCode: Setup & Get Free Frontier Models in 5 mins

TL;DR: This article covers the installation and configuration of the open-source AI coding agent OpenCode, focusing on how to use cutting-edge models for free via the built-in OpenCode Zen service, along with basic terminal interface operations. ## OpenCode Overview: An Open-Source, Provider-Agnostic AI Coding Agent Most popular AI coding agents today, such as Cloud Code (tied to Anthropic models) and Codex (tied to OpenAI models), are closed-source — users cannot see how they work internally. OpenCode, as an open-source alternative, is fully transparent: the tools it uses and the logic that coordinates the coding workflow are clearly visible. More importantly, OpenCode is provider-agnostic and supports Anthropic, OpenAI, Groq, and any other models. It even comes with its own free model service, OpenCode Zen, allowing you to use cutting-edge models without paying a cent. ## Installation: One Command Does It All OpenCode is very lightweight and installs with a single command. However, it requires a modern terminal (e.g., Ghosty, Alacritty, Kitty). Copy and paste the following command into your terminal to complete the installation: ```bash # The actual command depends on the official docs; the video does not show a specific command ``` After installation, OpenCode is added to your PATH. Refresh your terminal configuration (e.g., `source ~/.bashrc`), navigate to your working directory, and run `open-code` to launch the interface. ## Configuring Free Models: Connect to OpenCode Zen After launching, select "Connect to a provider" and it's recommended to use OpenCode Zen (free model). Press Enter, then you'll need to visit a provided URL to obtain an API key. Log in to the OpenCode Zen website; you can enable billing (not needed for free models) — just keep the balance at $0. Visit Zen to see available models: all models are listed, some with a "Free" label. You can filter to show only free models (about 6) or disable other models to restrict usage. Choose a free model, e.g., the default "big pickle" (which runs some open-source model, details not disclosed). Copy the generated API key, return to the terminal, paste it, and submit to complete the connection. Free models provided by OpenCode Zen include: MIMO, Nemotron 3 Ultra, DeepSeek V4 Flash, and others. These models often rank in the top 5 to top 10 on open-source model leaderboards and perform excellently. If you're willing to pay, you can also use Claude or GPT-5 Max subscriptions, but the free models are already powerful enough. ## First Look at the Interface and Basic Operations After connecting, test it: type any request, and the model responds normally. The right side of the interface shows token count, cost, whether LSP is available, etc. By default, the sidebar displays these details. If you find it distracting, you can hide it: - Press `Ctrl+X` then `B`, or - Press `Ctrl+P` to open the command list and search for "sidebar" to toggle its visibility. Another useful command is `/thinking`: when enabled, it expands to show the model's reasoning process, helping you observe its logic and correct errors promptly. You can enable or disable it as you prefer. ## Next Steps: Advanced Developer Workflows This video only covers the basic setup and interface of OpenCode. The next video will dive into advanced workflows: how to start a project (e.g., building a shooting game), how to iterate multiple times for different features, how to clear sessions to start new features, and so on. Stay tuned for the next installment. Source: OpenCode: Setup & Get Free Frontier Models in 5 mins (https://youtu.be/AdZ-WKvQels)