# Groq

> GroqCloud runs open-weight LLMs on custom LPU hardware for very fast, low-latency inference through an OpenAI-compatible API.

Groq is a hosted inference service that serves open-weight models — Llama, Qwen, Mistral, OpenAI open-weights, Whisper — on its custom LPU chips for unusually high tokens-per-second and low latency. The GroqCloud API is OpenAI-compatible, so most SDKs work by changing the base URL, and pricing is per-token with a free tier.

Website: https://groq.com

Groq is an AI inference company built around custom silicon. Where most providers serve models on GPUs, Groq designed the **LPU (Language Processing Unit)** specifically for inference, and the result is unusually high tokens-per-second and low time-to-first-token. Its hosted platform, **GroqCloud**, exposes that hardware through a simple API so you don't manage any of it yourself.

The catalog is **open-weight models** — Llama, Qwen, Mistral, OpenAI open-weights (GPT-OSS), and Whisper for speech-to-text — rather than closed frontier models. The draw is speed and predictable per-token pricing on those open models, not exclusive access to a proprietary model family.

The API is **OpenAI-compatible**: point an OpenAI client at `https://api.groq.com/openai/v1` with a Groq key and most existing code, SDKs, and tooling work with a one-line change. There is a usable free tier for prototyping, then per-token Developer pricing and Enterprise plans.

> [!NOTE]
> Groq (the LPU inference company) is unrelated to Grok, the xAI chatbot — they're easy to confuse but different products.

Groq sits alongside other hosted open-model inference services like [Together AI](/tools/together-ai), [Fireworks AI](/tools/fireworks-ai), and [Baseten](/tools/baseten); its differentiator is latency and throughput from the LPU. If you're weighing a hosted API against running models yourself, see [self-host vs API](/guides/architecture/self-host-vs-api-llm) and [LLM cost and latency engineering](/guides/architecture/llm-cost-latency-engineering).

---

_Source: https://agentscamp.com/tools/groq — Tool on AgentsCamp._