---

name: groq description: Expert guidance for Groq, the LLM inference platform that provides the fastest token generation speeds available, powered by custom LPU (Language Processing Unit) hardware. Helps developers integrate Groq's API for real-time AI applications where latency matters — chatbots, code completion, and streaming responses. license: Apache-2.0 compatibility: No special requirements metadata: author: terminal-skills version: 1.0.0 category: data-ai tags:

• llm
• inference
• api
• fast-inference
• chat

---

Groq — Ultra-Fast LLM Inference

Overview

Groq, the LLM inference platform that provides the fastest token generation speeds available, powered by custom LPU (Language Processing Unit) hardware. Helps developers integrate Groq's API for real-time AI applications where latency matters — chatbots, code completion, and streaming responses.

Instructions

Basic Chat Completion

// src/llm/groq-client.ts — Groq API (OpenAI-compatible)
import Groq from "groq-sdk";

const groq = new Groq({
  apiKey: process.env.GROQ_API_KEY!,
});

// Basic completion — fastest LLM inference available
async function chat(prompt: string): Promise<string> {
  const completion = await groq.chat.completions.create({
    model: "llama-3.3-70b-versatile",     // Fast and capable
    messages: [
      {
        role: "system",
        content: "You are a helpful assistant. Be concise and direct.",
      },
      { role: "user", content: prompt },
    ],
    temperature: 0.7,
    max_tokens: 1024,
  });

  return completion.choices[0].message.content ?? "";
}

// Streaming for real-time UI
async function streamChat(
  prompt: string,
  onChunk: (text: string) => void
): Promise<string> {
  const stream = await groq.chat.completions.create({
    model: "llama-3.3-70b-versatile",
    messages: [{ role: "user", content: prompt }],
    stream: true,
  });

  let fullResponse = "";
  for await (const chunk of stream) {
    const text = chunk.choices[0]?.delta?.content ?? "";
    fullResponse += text;
    onChunk(text);
  }
  return fullResponse;
}

Structured Output (JSON Mode)

// src/llm/structured.ts — Get structured JSON responses from Groq
async function extractEntities(text: string) {
  const completion = await groq.chat.completions.create({
    model: "llama-3.3-70b-versatile",
    messages: [
      {
        role: "system",
        content: `Extract entities from the text. Return JSON with this structure:
          { "people": [string], "organizations": [string], "locations": [string], "dates": [string] }`,
      },
      { role: "user", content: text },
    ],
    response_format: { type: "json_object" },
    temperature: 0,                        // Deterministic for extraction
  });

  return JSON.parse(completion.choices[0].message.content!);
}

// Tool use / Function calling
async function chatWithTools(prompt: string, tools: any[]) {
  const completion = await groq.chat.completions.create({
    model: "llama-3.3-70b-versatile",
    messages: [{ role: "user", content: prompt }],
    tools: [
      {
        type: "function",
        function: {
          name: "get_weather",
          description: "Get the current weather for a city",
          parameters: {
            type: "object",
            properties: {
              city: { type: "string", description: "City name" },
              unit: { type: "string", enum: ["celsius", "fahrenheit"] },
            },
            required: ["city"],
          },
        },
      },
      {
        type: "function",
        function: {
          name: "search_web",
          description: "Search the web for information",
          parameters: {
            type: "object",
            properties: {
              query: { type: "string", description: "Search query" },
            },
            required: ["query"],
          },
        },
      },
    ],
    tool_choice: "auto",
  });

  const toolCalls = completion.choices[0].message.tool_calls;
  if (toolCalls) {
    for (const call of toolCalls) {
      console.log(`Tool: ${call.function.name}, Args: ${call.function.arguments}`);
    }
  }
  return completion;
}

Audio Transcription (Whisper)

// src/audio/transcribe.ts — Fast audio transcription via Groq
import fs from "fs";

async function transcribeAudio(filePath: string) {
  const transcription = await groq.audio.transcriptions.create({
    file: fs.createReadStream(filePath),
    model: "whisper-large-v3-turbo",       // Fastest Whisper model
    language: "en",                         // Optional: specify language
    response_format: "verbose_json",       // Includes timestamps
    temperature: 0,
  });

  return {
    text: transcription.text,
    segments: transcription.segments?.map((s: any) => ({
      start: s.start,
      end: s.end,
      text: s.text,
    })),
    duration: transcription.duration,
  };
}

// Translate audio to English
async function translateAudio(filePath: string) {
  const translation = await groq.audio.translations.create({
    file: fs.createReadStream(filePath),
    model: "whisper-large-v3-turbo",
    response_format: "text",
  });
  return translation;
}

Model Selection

// Available Groq models (as of early 2026)
const MODELS = {
  // Llama 3.3 — best balance of speed and quality
  "llama-3.3-70b-versatile": {
    contextWindow: 128_000,
    bestFor: "general-purpose, reasoning, coding",
    speed: "~350 tok/s",
  },
  // Llama 3.1 — larger context, slightly slower
  "llama-3.1-8b-instant": {
    contextWindow: 128_000,
    bestFor: "simple tasks, classification, extraction",
    speed: "~750 tok/s",
  },
  // Mixtral — great for multilingual
  "mixtral-8x7b-32768": {
    contextWindow: 32_768,
    bestFor: "multilingual, creative writing",
    speed: "~500 tok/s",
  },
  // Gemma 2 — compact and efficient
  "gemma2-9b-it": {
    contextWindow: 8_192,
    bestFor: "lightweight tasks, edge deployment comparison",
    speed: "~600 tok/s",
  },
};

// Choose model based on task
function selectModel(task: "chat" | "extract" | "code" | "fast"): string {
  switch (task) {
    case "chat": return "llama-3.3-70b-versatile";
    case "extract": return "llama-3.1-8b-instant";     // Fast, good enough for extraction
    case "code": return "llama-3.3-70b-versatile";
    case "fast": return "llama-3.1-8b-instant";         // Lowest latency
  }
}

Python Integration

# src/groq_client.py — Groq with Python (OpenAI-compatible)
from groq import Groq

client = Groq(api_key=os.environ["GROQ_API_KEY"])

# Basic completion
response = client.chat.completions.create(
    model="llama-3.3-70b-versatile",
    messages=[{"role": "user", "content": "Explain quantum computing in 3 sentences"}],
    temperature=0.7,
    max_tokens=200,
)
print(response.choices[0].message.content)

# With OpenAI SDK (drop-in replacement)
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["GROQ_API_KEY"],
    base_url="https://api.groq.com/openai/v1",
)
# Same API, Groq's speed

Installation

# TypeScript/JavaScript
npm install groq-sdk

# Python
pip install groq

# Or use with any OpenAI-compatible SDK
# Just set base_url to https://api.groq.com/openai/v1

Examples

Example 1: Integrating Groq into an existing application

User request:

Add Groq to my Next.js app for the AI chat feature. I want streaming responses.

The agent installs the SDK, creates an API route that initializes the Groq client, configures streaming, selects an appropriate model, and wires up the frontend to consume the stream. It handles error cases and sets up proper environment variable management for the API key.

Example 2: Optimizing structured output performance

User request:

My Groq calls are slow and expensive. Help me optimize the setup.

The agent reviews the current implementation, identifies issues (wrong model selection, missing caching, inefficient prompting, no batching), and applies optimizations specific to Groq's capabilities — adjusting model parameters, adding response caching, and implementing retry logic with exponential backoff.

Guidelines

1. Use Groq for latency-sensitive tasks — Chatbots, autocomplete, real-time analysis; Groq's speed is 5-10x faster than cloud GPU providers
2. llama-3.1-8b for simple tasks — Don't use the 70B model for classification or extraction; 8B is faster and cheaper
3. JSON mode for structured output — Use response_format: { type: "json_object" } for reliable structured responses
4. Stream everything user-facing — With Groq's speed, streaming feels instant; still use it for perceived responsiveness
5. Whisper for audio — Groq's Whisper is the fastest transcription API available; process audio files in seconds
6. Rate limits vary by model — Check your rate limits in the Groq dashboard; 8B models have higher request limits
7. OpenAI SDK compatibility — Switch from OpenAI to Groq by changing the base URL; no code changes needed
8. Fallback to other providers — Groq can have capacity constraints during peak; have a fallback to OpenAI or Anthropic