50 models7 providersprices scraped nightly — no estimates

Groq

Verified May 27, 2026
7 models hosted · from $0.0500/1M

Groq runs open-weights models on its in-house LPU inference hardware, which is built for low-latency token generation rather than raw throughput. On supported models — currently a curated set of Llama 3.x, Mixtral, Qwen, and DeepSeek variants — observed tokens-per-second routinely beat GPU-based providers by 3–10×, with median time-to-first-token often under 200ms.

Go to Groq

Pricing is per-token and roughly competitive on input but higher on output than commodity GPU providers, reflecting the latency premium. The trade-off is catalog depth: new models land later than on Together or Fireworks, and quantization variants are limited.

Inference runs from US regions; enterprise plans add private clusters, but EU residency is not currently offered as a self-serve option.

groq.comPricing page ↗verified May 27, 2026

Calculate cost for your workload

Plug in your monthly tokens — get the actual bill on every provider serving each model.

Open calculator