opencode with an uncensored model: config

opencode.json

Save this as opencode.json in your project, or as ~/.config/opencode/opencode.json to use it everywhere. Put your own key in.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "uncensoredai": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Uncensored Chats",
      "options": {
        "baseURL": "https://uncensoredchats.com/v1",
        "apiKey": "sk-uc-..."
      },
      "models": {
        "glm-5.3-flash-xploded": { "name": "GLM 5.3 Flash" }
      }
    }
  },
  "model": "uncensoredai/glm-5.3-flash-xploded"
}

Start opencode and it will use the model named on the last line.

Why use an uncensored model for coding

Not for the code itself, most of the time. It matters when the work is security testing, scraping, reverse engineering or anything else a mainstream model declines half way through a task, leaving the agent stuck in a loop of apologies. A model that does not refuse finishes the job or fails with a real error.

Which model

GLM 5.3 Flash is the default pick for an agent: quick, cheap, good at tool calls, a million tokens of context.
GLM 5.3 (glm-5.3-xploded) when the task needs more careful reasoning. Text only.
Kimi K3.1 Pro (kimi-k3.1-pro-xploded) for the hardest work, at a much higher output price.

Cap the key first

A coding agent can send hundreds of requests on its own. In the dashboard, give the key you use here a total or monthly spend cap. If the agent runs away, the key stops and the rest of your balance is untouched.

What it costs

Agents resend the same files and history on every step, and that repeated input is billed at the cached price: $0.20 per million tokens on GLM 5.3 Flash against $0.30 for new input.

Keep reading

Updated 12 November 2026. Prices are read from the live price list.