Skip to content

GitHub Copilot Setup Guide

Aditya Katre edited this page Nov 9, 2026 · 7 revisions

OCS Intelligence currently supports three models, but we are currently working on adding faster, intelligent models.

  • Qwen3.8:27b - Most intelligent, but typically overthinks. Best for difficult, time-insensitive tasks.
  • Gemma4:26b - Best for tool calling and agentic tasks. Faster than qwen3.8:27b. MoE.
  • **Qwen3.6:35b-a3b **- Best for general-purpose usage. Faster than Qwen3.8:27b. Best for long-horizon tasks. MoE.

Please complete the steps outlined below to use OCS models within GitHub Copilot Chat, free of charge.

First, click the plus in Copilot chat:

image

Then, press Add Models -> Custom Endpoint:

image

Choose any name for the provider (ocs_intelligence is fine)

Please request an API Key from the maintainers of OCS Intelligence

Paste the following to add our models, following standard JSON conventions:

	{
		"name": "ocs_intelligence",
		"vendor": "customendpoint",
		"apiKey": "${input:chat.lm.secret.-3581f028}",
		"apiType": "chat-completions",
		"models": [
			{
				"id": "qwen3.8:27b",
				"name": "qwen3.8:27b",
				"url": "https://ai.opencodingsociety.com/v1",
				"toolCalling": true,
				"vision": true,
				"maxInputTokens": 163840,
				"maxOutputTokens": 16384
			},
			{
				"id": "gemma4:26b",
				"name": "gemma4:26b",
				"url": "https://ai.opencodingsociety.com/v1",
				"toolCalling": true,
				"vision": true,
				"maxInputTokens": 256000,
				"maxOutputTokens": 16384
			},
			{
				"id": "qwen3.6:35b-a3b",
				"name": "qwen3.6:35b-a3b",
				"url": "https://ai.opencodingsociety.com/v1",
				"toolCalling": true,
				"vision": true,
				"maxInputTokens": 128000,
				"maxOutputTokens": 16000
			}
		]
	}

Clone this wiki locally