Groq - AI Tools Nav

What it is for

Groq is a hosted inference API on GroqCloud: pick a model ID, then call the OpenAI-shaped /openai/v1 endpoint. Use it when the job is fast token generation billed per 1M tokens.

What makes it unique: the cover is the live Supported Models table — Featured GPT-OSS 120B plus printed PRICE PER 1M TOKENS rows.

Key features

Models catalog — one Supported Models table for chat, Whisper, TTS, guard, and Compound IDs, with T/sec and price-per-1M columns (cover) on Supported Models.
Chat Completions — send a model ID from that table to https://api.groq.com/openai/v1 (Quickstart).
Speech-to-text — Whisper large-v3 / turbo transcription on the Speech to Text endpoint.
Tool use — function calling, Groq built-in tools (browser search / code execution), or remote MCP in one API loop (Tool Use).
Vision — image URL or base64 on multimodal chat models for VQA / caption / OCR (Images and Vision).

Who it's for

Best fit: developers who will copy a model ID and a $/1M row from the Supported Models table.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from the live Supported Models table. Copy the row you will call.

Production examples (price per 1M tokens, Developer plan):

GPT-OSS 120B (openai/gpt-oss-120b) — input $0.15 / output $0.60. ~500 T/sec. 250K TPM / 1K RPM.
GPT-OSS 20B (openai/gpt-oss-20b) — input $0.075 / output $0.30. ~1000 T/sec. 250K TPM / 1K RPM.
Whisper large-v3 — $0.111 per hour.
Whisper large-v3-turbo — $0.04 per hour.

Preview examples on the same page:

Qwen3.8-27B — input $0.80 / output $4.00.
Llama Prompt Guard 2 22M — $0.03 / $0.03.
Orpheus V1 English — $22.00 per 1M characters.

Llama 3.1 8B Instant and Llama 3.3 70B Versatile print Contact Sales.

Limitations / costs

Rows move. Pin a model ID if you need a stable price.
Enterprise Llama rows are quote-only on the public table.
Preview models can retire on short notice (docs note on the same page).
Rate limits are organization-wide (RPM / TPM / RPD). Cached tokens stay outside those caps.

Alternatives

Together AI — hosted open-model catalog with printed $/1M cards.
OpenRouter — one key across many providers; Standard 5.5% credit fee.
Cerebras — Instant AI lobby plus public token rows.

Official tutorial

Written path: Supported Models plus Quickstart.

Website

https://groq.com

Source

Catalog row curated for AI Tools Nav. Cover is the live GroqCloud Supported Models table (GPT-OSS price rows), cropped past the cookie dialog. Dollars checked on console.groq.com/docs/models (GPT-OSS 120B $0.15 / $0.60; 20B $0.075 / $0.30; Whisper $0.111 / $0.04 per hour).

Last verified

2026-10-22 (Asia/Shanghai) editorial check. Only rows that printed $ or Contact Sales were copied.