Groq
Groq (Note: not Grok by xAI) is an inference service known for LPU hardware acceleration and ultra-low latency. It primarily hosts open-source models such as Llama, Mixtral, and Whisper, with output speeds typically several times faster than standard cloud inference.
Get API Key
- Go to GroqCloud to register an account
API Keys→Create API Key, copy thegsk_...key
Configure in BLACK HOLE
- Open
Settings → Model Provider, find the Groq Provider and enter the details page - Enter your
gsk_...in the API Key field - The API Endpoint defaults to
https://api.groq.com/openai/v1; no modification is needed - Click Fetch Model List
Recommended Usage
| Model | Suitable Scenarios |
|---|---|
llama-3.3-70b-versatile | General conversation, extremely fast |
llama-3.1-8b-instant | Simple tasks, millisecond-level response |
mixtral-8x7b-32768 | Long context |
whisper-large-v3 | Speech-to-text |
Use Cases
- Real-time chatbots: Groq's "near-instant response" is ideal for IM integration (paired with Channels)
- High concurrency: Tokens per second are significantly higher than standard cloud inference
- Not concerned with the latest models: Groq primarily hosts open-source models like the Llama series; it does not offer closed-source models like GPT-5 or Claude-4
Distinguishing Grok vs Groq
| Grok | Groq | |
|---|---|---|
| Company | xAI (Musk) | Groq Inc. |
| Focus | Proprietary LLMs + Web access | LPU hardware + Open-source model inference |
| Models | grok-4 and other proprietary models | llama-3.x and other open-source models |
Get Help and Submit Feedback
If you have any questions, bugs, or feature improvement suggestions while configuring or using the service, please refer to the official channels provided in Feedback and Suggestions.