Enki

Groq review: GroqCloud API speed, pricing and limits

Fast inference API for open models, on its own LPU chips

8.0Editor’s scoreFreemiumfrom Pay as you goPricing verified Nov 17, 2026

Screenshots

GroqCloud

Fast inference for open models

8.0OUT OF 10

Enki editor score

Excellent

Human-vetted editorial rating

Earned by a no-card free plan, low per-token prices such as gpt-oss-120b at $0.15/$0.60 and a 50% Batch discount; cost by a short public model list and Llama models moving to Enterprise-only pricing.

We may earn a commission if you sign up through our links. It never affects our ratings or which stories we cover.

Overview

Groq runs GroqCloud, an inference cloud built on its own LPU chips, and sells API access to open models such as OpenAI's gpt-oss and Qwen at high output speeds. The endpoint is OpenAI-compatible, so most apps switch by changing the base URL and key. A free plan with tight rate limits needs no card; the Developer plan is pay per token, billed monthly in arrears, and adds higher limits, Batch and Flex processing and spend limits. Some models, including Llama 3.1 8B and Llama 3.3 70B, are now listed as Enterprise models with prices by contact only. In December 2025 Groq licensed its inference technology to NVIDIA on a non-exclusive basis and said GroqCloud would keep operating.

Enki verdict

Groq is best at serving open models fast through a cheap, OpenAI-compatible API with a usable free plan. Reach for it when speed per request matters for an app built on gpt-oss or Qwen, not for teams that need a wide catalog of closed models behind one key.

Key features

Fast inference

Open models served on Groq's LPU hardware at high token speeds.

OpenAI-compatible API

Point an existing OpenAI client at api.groq.com and swap the key.

Batch and Flex

Cheaper or more flexible processing tiers for the Developer plan.

Pros & cons

What we like

  • Free plan with no card needed
  • OpenAI-compatible, so switching is a base URL change
  • Batch API at half price

Worth noting

  • Short public model list
  • Llama models are now Enterprise-only, priced by contact
  • Free limits are tight at 8,000 tokens a minute

Groq pricing

GroqCloud has a free plan, a pay-per-token Developer plan and Enterprise contracts; prices are listed per model on Groq's models page.

Is Groq free?

Yes. Groq's Free plan gives API access to its public models with no card, capped by per-model rate limits.

  • gpt-oss-120b and gpt-oss-20b: 30 requests a minute, 1,000 a day, 8,000 tokens a minute and 200,000 tokens a day
  • qwen/qwen3.8-27b: 30 requests a minute and 1,000 a day
  • Rate limits apply per organization, not per user
  • No Batch or Flex processing on the Free plan
Groq plans and prices
DeveloperPay per tokenmonthly in arrears; early invoices at $1, $10, $100, $500 and $1,000 of lifetime usage
  • gpt-oss-120b: $0.15 input, $0.60 output per 1M tokens
  • gpt-oss-20b: $0.075 input, $0.30 output per 1M tokens
  • Batch API at a 50% discount, plus Flex processing and spend limits
  • Base limit for gpt-oss-120b: 1,000 requests and 250,000 tokens a minute
EnterpriseContact sales
  • Enterprise-only models such as Llama 3.1 8B, Llama 3.3 70B and MiniMax M2.7
  • Higher limits for select workloads

Source: console.groq.com/docs/models
Checked on: . Prices can change; confirm at checkout.

What it means for you

Start on the Free plan to test speed and model quality; 1,000 requests a day is enough for a prototype. Upgrade to Developer when you hit the limits: it costs nothing until you use tokens, and Batch halves the price of jobs that can wait. Groq's public lineup is short, so if you need many models behind one key, OpenRouter covers far more. Looking for xAI's chatbot instead? That is Grok. See the best AI coding tools.

Price history

  1. Checked against https://console.groq.com/docs/models (groq.com/pricing now redirects to the homepage): gpt-oss-120b $0.15/$0.60 and gpt-oss-20b $0.075/$0.30 per 1M tokens; Llama 3.1 8B and 3.3 70B listed as Enterprise, contact sales; Batch 50% off.

What real users think

  • “I use Whispering connected to the whisper-large-v3-turbo model on Groq. It's incredibly cheap and works reliably for me.”

  • “Also Groq is very fast, but the latency wasn't always consistent and I saw some very strange responses on a few calls that I had to attribute to quantization.”

  • “The problem with groq was they only allowed LORA on llama 8b and 70b, and you had to have an enterprise contract it wasn't self service.”

Quoted word for word from public posts, shortened only where marked with “…”. We pick a spread of views, not the kindest ones, and these comments do not affect the Enki score. Checked Nov 18, 2026.

Groq FAQ

Is Groq free?
Yes, within limits. The Free plan needs no payment method; for gpt-oss-120b it allows 30 requests a minute, 1,000 requests a day and 200,000 tokens a day. Higher limits, Batch and Flex need the pay-as-you-go Developer plan.
How much does Groq cost?
On the Developer plan you pay per token, by model. gpt-oss-120b is $0.15 per 1M input tokens and $0.60 per 1M output tokens, gpt-oss-20b is $0.075 and $0.30, and the Batch API is 50% cheaper. Some models, such as Llama 3.3 70B, are listed for Enterprise customers with prices by contact.
Is Groq the same as Grok?
No. Groq is an inference company that runs open models on its own chips through GroqCloud. Grok is the chatbot and model family made by Elon Musk's xAI, which Enki covers on its own page.
Is Groq still independent after the NVIDIA deal?
Groq says yes. In December 2025 it announced a non-exclusive licensing agreement for NVIDIA to use its inference technology, with founder Jonathan Ross and other staff moving to NVIDIA, and said Groq remains independent and GroqCloud would continue without interruption.

No written reviews yet — be the first to share your experience.

Related tools