Groq
Groq is an AI inference cloud built on its own LPU chips, with infrastructure, inference and controls in one integrated platform. It is built for developers and companies running AI models in production who need fast inference.
Your chatbot works, but users stare at a spinner while the answer trickles out. Groq is an inference cloud built to make that wait shorter. It runs open models on its own LPU chips and serves them through an API that speaks the same format as OpenAI's, so for many apps switching means changing a base URL and an API key.
The catalog covers text generation and reasoning, speech to text with Whisper, text to speech, image understanding and content moderation. There are built-in tools for code execution and browser search on the GPT OSS models, plus support for remote MCP tools, so agents can call out to other services.
The people signing up are mostly developers and AI teams who care about response speed and cost per token.
Pricing is per use. On the models page, GPT OSS 120B is $0.15 per million input tokens and $0.60 per million output tokens, GPT OSS 20B is $0.075 and $0.30, and Whisper Large V3 Turbo transcription is $0.04 per hour of audio. Some Llama models are listed as Enterprise, with pricing through sales. Prices checked on console.groq.com in October 2026.
Worth knowing before you sign up
- Rate limits apply to the whole organization, not to each person, and you can hit requests per minute before you hit tokens per minute.
- Cached tokens do not count toward your rate limits.
- Preview models are for evaluation only and may be withdrawn, so build production apps on the production list.
For developers
There are batch and flex processing tiers for jobs that can wait, spend limits and projects in the console, LoRA inference, and ready-made connectors for Gmail, Google Calendar and Google Drive. Coding agents such as Cline, Roo Code and Kilo Code can use Groq as their model provider.
twenty47tools does not endorse or recommend any of the tools you find in our directory.