Groq on Hugging Face Inference Providers 🔥

Source: HuggingFace Date: 2025-06-16 URL: https://huggingface.co/blog/inference-providers-groq

Summary

Integration announcement: Groq is now an Inference Provider on the HuggingFace Hub. Users can route inference for open-source models (Llama 4, Qwen QWQ-32B) through Groq’s LPU hardware either via their own Groq API key or billed through HF accounts. HF PRO users get $2/month in inference credits. Access is via the hub UI or the huggingface_hub Python/JS SDK with provider="groq".

Implications

HF as open-source ML hub. Each new inference provider added to the Hub tightens HF’s position as the single routing layer for open-model inference — similar to how cloud providers centralize compute. Groq’s LPUs are specifically pitched on low-latency, which is distinct from the throughput-optimized providers already available.

Open-weights ecosystem health. Groq’s LPU becoming accessible without a direct Groq account lowers the barrier to high-speed open-weight inference, particularly for developers already inside the HF ecosystem. Watch whether latency-sensitive use cases (agents, real-time) migrate toward LPU-backed providers once pricing is comparable.