Troubleshoot devices and models
Check ShareAI provider eligibility, device connectivity, model requirements and remote access in a practical troubleshooting order.
On this page
Start with the observed state#
Record the provider workspace, device name, exact model identifier and time of the problem. Then find the first failing step below. Avoid changing several settings at once; a before-and-after comparison is easier to understand.
| Problem | Check first | Then check |
|---|---|---|
| Device missing | Correct provider and workspace. | Completed sign-in/connection and the device’s current network. |
| Connected, no sharing | Eligible model installed and enabled. | Model loading errors and available compute. |
| Model cannot load | Exact variant and current hardware requirements. | GPU memory used by other processes, storage, drivers and runtime logs. |
| Remote API returns 403 | Correct credential type, provider UUID and device UUID. | Both enabled switches, token status and connected/sharing relationship. |
| Command accepted, state unchanged | Actual device connection and command result. | Installed/enabled models and subsequent session activity. |
| No tasks | Observed availability and eligibility. | Current demand and model access; being online does not guarantee work. |
Hardware requirements are model-specific#
Use the requirements shown for the selected model and application version. GPU memory demand changes with quantization, context and concurrency. Historical parameter-to-VRAM estimates do not establish current eligibility.
Prepare useful diagnostics#
- Exact application/runtime version and operating system.
- GPU and memory information relevant to the failing model.
- Provider/device/model identifiers without secrets.
- Time, HTTP status, error code and task/session identifier when available.
Where to continue#
Contact Support or revisit device access requirements.
Last updated September 15, 2026