Groq
Understand GroqCloud inference, how it differs from Hub discovery, and its current status in hugpad.
7 sections / hugpad handbook
On this page
Current integration status
Not connected in this version
What Groq provides
Groq offers hosted inference through GroqCloud. Instead of browsing a repository, an inference request sends input to a supported running model and receives an output.
The models available on Groq are a provider-specific selection. Finding a repository on Hugging Face does not establish that the same model, revision, or feature set is supported by Groq.
Hugging Face and Groq have different roles
| Service | Role | Status in hugpad |
|---|---|---|
| Hugging Face Hub | Repository discovery, model cards, publisher profiles, and identity. | Live Hub connection and sign-in. |
| GroqCloud | Hosted execution of supported models through an inference API. | Reference documentation only; not connected. |
| pump.fun / Solana | Token creation and transaction execution, not AI inference. | Wallet-approved launch flow; success requires confirmation. |
Supported models and capabilities
Use Groq’s supported-model list for current model IDs, production/preview status, and model-specific capabilities. Do not assume a provider serves every public Hub model. Availability, context limits, and supported features can change; this guide deliberately does not hardcode an unverified model list.
API compatibility and credentials
Groq documents an API that is mostly compatible with OpenAI client libraries, with differences and unsupported features documented separately. Compatibility is not a guarantee that every option or workflow works unchanged.
An application integration needs a valid Groq account, server-held credentials, a supported model, and deliberate handling of errors and limits. Never paste an API key into public documentation, a token description, or client-visible code. See the official quickstart and compatibility guide.
Usage, limits, and privacy
Usage and rate limits depend on the account, model, and request type. Review your actual Groq account and rate-limit documentation rather than relying on a universal allowance or an unverified latency claim.
A hosted inference request sends its input to that provider. Review applicable terms and data handling before sending private information. Browsing hugpad does not currently send prompts or model-card content to Groq.