ORYNX: one API for every frontier model

OpenCode can mix several providers. Use the OpenAI-compatible provider for Kimi, GLM, DeepSeek and Grok, the Anthropic provider for Claude, and the OpenAI provider for GPT (it calls the Responses API).

Put this in opencode.json in your project, or in ~/.config/opencode/opencode.json to use it in every project:

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "orynx": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "ORYNX",
      "options": { "baseURL": "https://api.orynx.dev/v1", "apiKey": "{env:ORYNX_API_KEY}" },
      "models": {
        "kimi-k3": { "name": "Kimi K3", "limit": { "context": 1048576, "output": 131072 } },
        "glm-5.3": { "name": "GLM-5.3", "limit": { "context": 1048576, "output": 131072 } }
      }
    },
    "orynx-claude": {
      "npm": "@ai-sdk/anthropic",
      "name": "ORYNX Claude",
      "options": { "baseURL": "https://api.orynx.dev/v1", "apiKey": "{env:ORYNX_API_KEY}" },
      "models": {
        "claude-sonnet-5": { "name": "Claude Sonnet 5", "limit": { "context": 1000000, "output": 128000 } },
        "claude-opus-5": { "name": "Claude Opus 5", "limit": { "context": 1000000, "output": 128000 } }
      }
    },
    "orynx-gpt": {
      "npm": "@ai-sdk/openai",
      "name": "ORYNX GPT",
      "options": { "baseURL": "https://api.orynx.dev/v1", "apiKey": "{env:ORYNX_API_KEY}" },
      "models": {
        "gpt-5.5": { "name": "GPT-5.5", "limit": { "context": 1050000, "output": 128000 } }
      }
    }
  }
}

export ORYNX_API_KEY="sk-orynx-..."
opencode

Run /models and pick a model, for example orynx-claude/claude-sonnet-5. limit is the model's context window and output cap. Each model's page under Models lists both.

The Vercel AI SDK Anthropic provider (@ai-sdk/anthropic) expects /v1 in its base URL, unlike the official Anthropic SDKs.