Run NVIDIA Kimi K2.6 and others in OpenCode
I wanted to use Kimi K2.6 from NVIDIA inside OpenCode, and it was simpler than I first thought. This and a bunch of other open-source models are free to try in NVIDIA, by the way.
I only had to add the API key, model details, and the correct base URL in the OpenCode config.
I wanted to try Kimi K2.6 but you can try other models as well. For Kimi, NVIDIA's model name is:
moonshotai/kimi-k2.6
But inside OpenCode, I had use it like below
nvidia/moonshotai/kimi-k2.6
I edited this file on my macOS device:
~/.config/opencode/opencode.jsonc
And added this config:
{
"$schema": "https://opencode.ai/config.json",
"model": "nvidia/moonshotai/kimi-k2.6",
"provider": {
"nvidia": {
"npm": "@ai-sdk/openai-compatible",
"name": "NVIDIA",
"api": "https://integrate.api.nvidia.com/v1",
"options": {
"baseURL": "https://integrate.api.nvidia.com/v1",
"apiKey": "YOUR_NVIDIA_API_KEY_HERE",
"timeout": 600000,
"chunkTimeout": 60000,
"extraBody": {
"chat_template_kwargs": {
"thinking": true
}
}
},
"models": {
"moonshotai/kimi-k2.6": {
"name": "Kimi K2.6",
"reasoning": true,
"tool_call": true,
"temperature": true,
"attachment": true,
"interleaved": {
"field": "reasoning_content"
},
"modalities": {
"input": ["text", "image", "video"],
"output": ["text"]
},
"limit": {
"context": 262144,
"output": 65536
},
"options": {
"chat_template_kwargs": {
"thinking": true
}
}
}
}
}
}
}
The important bit is this URL:
https://integrate.api.nvidia.com/v1
NVIDIA's example uses the full /chat/completions URL, but OpenCode only needs the base URL. It adds /chat/completions by itself.
Works as expected, so that's it.
Webmentions