GMI Cloud Provider: OpenAI-Compatible Hosted Inference for OpenClaw
Learn how to use GMI Cloud's OpenAI-compatible API with OpenClaw, covering installation, auth, and model references. Ideal for developers needing a single key for multiple hosted model families.
Read this when
- You want to run OpenClaw with GMI Cloud models
- You need the GMI provider id, key, or endpoint
GMI Cloud offers a hosted inference service for both frontier and open-weight models, exposing an API that is compatible with OpenAI's. Within OpenClaw, it ships as an official external provider plugin: a single installation, credential storage through the standard model auth flow, and model references such as gmi/openai/gpt-5.6-sol are all that is needed.
Opt for GMI when a single API key should cover multiple hosted model families, among them the Anthropic, DeepSeek, Google, Moonshot, OpenAI, and Z.AI routes that GMI's catalog exposes. It suits scenarios where a secondary provider is needed for fallback, where hosted routes from different vendors are being compared, or where GMI lists a model ahead of your main provider. OpenClaw manages the provider id, auth profile, aliases, model catalog seed, and base URL; GMI handles live model availability, billing, rate limits, and any routing policy on the provider side.
| Property | Value |
|---|---|
| Provider id | gmi (aliases: gmi-cloud, gmicloud) |
| Package | @openclaw/gmi-provider |
| Auth env var | GMI_API_KEY |
| API | OpenAI-compatible (openai-completions) |
| Base URL | https://api.gmi-serving.com/v1 |
| Default model | gmi/openai/gpt-5.6-sol |
Setup
After installing the plugin and restarting the gateway, generate an API key within GMI Cloud (https://www.gmicloud.ai/):
openclaw plugins install @openclaw/gmi-provider
openclaw gateway restart
Next, execute:
openclaw onboard --auth-choice gmi-api-key
For non-interactive environments, --gmi-api-key <key> can be passed, or the following can be set:
export GMI_API_KEY="<your-gmi-api-key>" # pragma: allowlist secret
When to choose GMI
- A hosted OpenAI-compatible endpoint is preferred over running a local model server.
- Multiple commercial and open-weight model families should be testable through one provider account.
- A fallback provider with different upstream routing than DeepInfra, OpenRouter, Together, or the direct vendor APIs is required.
- GMI-specific model ids, pricing, or account controls are needed.
When vendor-native features that GMI does not expose via its OpenAI-compatible route are required, pick the direct vendor provider instead. If data locality or local GPU control outweighs hosted convenience, choose a local provider like LM Studio, Ollama, SGLang, or vLLM.
Models
The plugin catalog seeds route ids commonly available on GMI Cloud:
| Model ref | Input | Context | Max output |
|---|---|---|---|
gmi/anthropic/claude-sonnet-5 | text + image | 409,600 | 128,000 |
gmi/deepseek-ai/DeepSeek-V4-Pro | text | 1,048,576 | 384,000 |
gmi/google/gemini-3.5-flash-lite | text + image | 1,048,576 | 65,536 |
gmi/openai/gpt-5.6-sol | text + image | 1,050,000 | 128,000 |
gmi/zai-org/GLM-5.2-FP8 | text | 1,048,576 | 128,000 |
This catalog acts only as a starting point, not a guarantee that every model is callable on every account at any moment. To see what the configured provider reports in your environment, list it:
openclaw models list --provider gmi
Troubleshooting
- With
401or403: verify thatGMI_API_KEYis set for the process running OpenClaw, or rerun onboarding to save the key in the provider auth profile. - For unknown model errors: check that the model exists in your GMI account and use the full
gmi/<route-id>ref as shown byopenclaw models list --provider gmi. - For intermittent provider errors: switch to a different GMI route or set GMI as a fallback instead of the sole primary model provider.