Groq API Not Working? Fix Auth, Rate Limits & Model Errors

Authentication — 401 error

A 401 error means the API key is missing, invalid, or passed incorrectly. Use the Groq SDK or the OpenAI SDK pointed at Groq's base URL:

# Python — using groq SDK
from groq import Groq
import os

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

response = client.chat.completions.create(
    model="llama-3.3-70b-versatile",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

# Or using OpenAI SDK with Groq base URL
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["GROQ_API_KEY"],
    base_url="https://api.groq.com/openai/v1"
)

• Get API key at console.groq.com — go to API Keys → Create API Key. Keys start with gsk_.
• Set as env var GROQ_API_KEY — or pass via Authorization: Bearer header.
• Legacy key format — if your key doesn't start with gsk_, it may be an old format; generate a new key.
• Playground key is separate — Groq's playground uses its own session token, not an API key. Always use keys from the API Keys section.

Rate limits — free tier

Free tier limits apply per model, not globally — using different models doesn't share the same bucket:

Strategy: rotate between models to effectively multiply your rate limit:

import itertools
import time

MODELS = ["llama-3.3-70b-versatile", "llama-3.1-8b-instant", "mixtral-8x7b-32768"]
model_cycle = itertools.cycle(MODELS)

def groq_call(client, prompt):
    for attempt in range(5):
        try:
            return client.chat.completions.create(
                model=next(model_cycle),
                messages=[{"role": "user", "content": prompt}]
            )
        except Exception as e:
            if "429" in str(e) and attempt < 4:
                time.sleep((2 ** attempt) + 0.5)
            else:
                raise

• Upgrade to paid plan for higher limits at console.groq.com.
• Use request queuing rather than parallel requests to stay within the per-minute cap.

Current model names

Groq deprecates old model versions frequently. Always verify at console.groq.com/docs/models. As of June 2026:

Deprecated (will return 404): llama2-70b-4096, llama3-70b-8192 — use llama-3.3-70b-versatile instead.

Context length exceeded

Each Groq model has different context limits. If you hit the limit, switch to a model with a larger context window:

For conversation history management, trim older messages to stay within limits:

def trim_messages(messages, max_tokens=100000):
    # Keep system message + last N messages
    system = [m for m in messages if m["role"] == "system"]
    conversation = [m for m in messages if m["role"] != "system"]
    # Simple trim: keep system + last 20 exchanges
    return system + conversation[-40:]

Streaming not working

Groq uses the OpenAI-compatible API for streaming. Iterate chunks — do NOT call .content on the streaming response object directly:

# Python streaming
response = client.chat.completions.create(
    model="llama-3.3-70b-versatile",
    messages=[{"role": "user", "content": "Write a short story"}],
    stream=True
)

for chunk in response:
    content = chunk.choices[0].delta.content
    if content:
        print(content, end="", flush=True)

// JavaScript streaming
const stream = await client.chat.completions.create({
  model: "llama-3.3-70b-versatile",
  messages: [{ role: "user", content: "Write a story" }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

• Set stream=True (Python) or stream: true (REST/JS) in the request.
• REST response is Server-Sent Events — each data: line contains a JSON chunk.