Skip to content

Groq

Groq is a US AI inference company, founded in 2016 by Jonathan Ross, that makes the LPU chip and runs GroqCloud, a developer API for fast model inference. In December 2025 Nvidia signed a non-exclusive licensing agreement for Groq's inference technology, and Groq said it would continue as an independent company. Groq is not the same as xAI's Grok.

By GroqPay as you goSince 2016AI Platforms & APIs

Facts checked on against the sources listed at the end of this page.

Pricing
Pay as you goFree plan: Yes, with rate limits that vary by model. The paid Developer plan raises them.
Open source
NoProprietary
npm downloads a week
1.2Mdown 1% on the week before

A good fit for

  • Apps where response speed matters, such as voice or interactive agents
  • Developers who want to run open models through a hosted API
  • Teams that want an alternative inference provider to GPU clouds

Not a good fit for

  • Training models: it is an inference service
  • Running Groq hardware yourself: access is through the hosted service
  • Frontier proprietary models: you use the models Groq hosts

Groq pricing

GroqCloud has a free plan with rate limits that vary by model, and a paid Developer plan with higher limits plus Batch and Flex processing. Per-token prices are listed by model on Groq's pricing page.

PlanPriceWhat you get
Free$0Limited access with rate limits.
Usage basedPer token, by modelSee the pricing page for current rates.

Prices in USD, checked on 6 October 2026. Current prices on groq.com.

Main features

GroqCloud API
Developer platform for running models through an API.
LPU inference
Runs on Groq's own chips, designed for low-latency inference.
Pay as you go
Billed by usage; see the pricing page for current per-token rates.
Enterprise options
Contact Groq for dedicated capacity and custom terms.

About Groq

Groq designs the LPU, a chip built for AI inference, and sells access to it. GroqCloud, the developer platform, soft-launched in February 2024. The company describes itself as a neocloud for fast inference.

According to Wikipedia, in December 2025 Nvidia entered a non-exclusive licensing agreement for Groq's inference technology, reported at about $20 billion. It was a licensing arrangement and not an acquisition, and Jonathan Ross and COO Sunny Madra joined Nvidia. Groq said it would keep operating independently, and no GroqCloud closure has been announced.

Groq's homepage says its LPX technology works alongside Nvidia GPUs.

Made by
Groq
Launched
2016
Based in
Mountain View, California
Free plan
Yes, with rate limits that vary by model. The paid Developer plan raises them.
Available on
GroqCloud API, LPU hardware

Usage and activity

Weekly npm downloads of groq-sdk, last 12 months.

250k500k750k1M1.3MOctDecFebAprJunAugNov18 October 2025: 392,109 downloads a week25 October 2025: 364,126 downloads a week1 November 2025: 366,559 downloads a week8 November 2025: 397,688 downloads a week15 November 2025: 394,018 downloads a week22 November 2025: 414,768 downloads a week29 November 2025: 385,522 downloads a week6 December 2025: 432,021 downloads a week13 December 2025: 437,626 downloads a week20 December 2025: 426,970 downloads a week27 December 2025: 314,488 downloads a week3 January 2026: 305,717 downloads a week10 January 2026: 415,854 downloads a week17 January 2026: 442,257 downloads a week24 January 2026: 539,943 downloads a week31 January 2026: 469,815 downloads a week7 February 2026: 509,214 downloads a week14 February 2026: 601,777 downloads a week21 February 2026: 503,422 downloads a week28 February 2026: 587,780 downloads a week7 March 2026: 601,693 downloads a week14 March 2026: 627,363 downloads a week21 March 2026: 596,137 downloads a week28 March 2026: 665,918 downloads a week4 April 2026: 622,714 downloads a week11 April 2026: 635,726 downloads a week18 April 2026: 655,353 downloads a week25 April 2026: 677,630 downloads a week2 May 2026: 731,105 downloads a week9 May 2026: 742,258 downloads a week16 May 2026: 777,778 downloads a week23 May 2026: 758,590 downloads a week30 May 2026: 728,055 downloads a week6 June 2026: 769,561 downloads a week13 June 2026: 918,300 downloads a week20 June 2026: 979,820 downloads a week27 June 2026: 889,592 downloads a week4 July 2026: 793,545 downloads a week11 July 2026: 837,261 downloads a week18 July 2026: 887,071 downloads a week25 July 2026: 882,396 downloads a week1 August 2026: 856,981 downloads a week8 August 2026: 877,786 downloads a week15 August 2026: 954,373 downloads a week22 August 2026: 953,239 downloads a week29 August 2026: 938,821 downloads a week5 September 2026: 1,011,190 downloads a week27 September 2026: 990,016 downloads a week4 October 2026: 1,059,795 downloads a week31 October 2026: 1,233,829 downloads a week7 November 2026: 1,221,904 downloads a week14 November 2026: 1,207,814 downloads a week

Groq alternatives

All 15 alternatives to Groq, side by side

Questions about Groq

Did Nvidia acquire Groq?

No. According to Wikipedia, Nvidia signed a non-exclusive licensing agreement for Groq's inference technology in December 2025, and Groq said it would continue as an independent company.

Is Groq the same as Grok?

No. Groq is an inference company and chip maker. Grok is xAI's model.

Does Groq have a free tier?

Yes, GroqCloud offers a free tier with rate limits. Current limits are on the pricing page.

Who founded Groq?

Jonathan Ross founded Groq in 2016. It is based in Mountain View, California.

What is GroqCloud?

Groq's developer platform for running models on its LPU chips through an API. It soft-launched in February 2024.

Sources and links

Work on Groq? Claim this listing. Something wrong or out of date? Suggest a correction. Stats come from the public GitHub and npm APIs and update every day.

Groq: pricing, features and alternatives (2026) | ScratchDB