OpenAI API Not Working? Fix Authentication, Rate Limits & SDK Errors

401 Unauthorized / Invalid API key

The most common cause is a missing, expired, or incorrectly passed API key. Use the official SDK with an environment variable:

# Python
from openai import OpenAI
import os

client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)

// Node.js
import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

• Key format: standard keys start with sk-; project-scoped keys start with sk-proj- — both are valid.
• Check expiry: go to platform.openai.com/api-keys — deleted or expired keys show as inactive.
• Organization ID: if your account belongs to an organization, verify the org ID matches the key's organization in the API dashboard.
• Trial credit: free trial API keys expire after 3 months of inactivity or when the $5 trial credit runs out — add a payment method to continue.

429 Rate Limit — RPM/TPM vs billing limit

There are two distinct 429 errors. RPM/TPM limit: too many requests or tokens per minute — back off and retry. Billing limit: monthly spend cap reached — raise the limit at platform.openai.com/settings/billing. Add exponential backoff for RPM errors:

import time
import random
from openai import RateLimitError

def completion_with_retry(client, **kwargs):
    for attempt in range(5):
        try:
            return client.chat.completions.create(**kwargs)
        except RateLimitError as e:
            if attempt == 4:
                raise
            wait = (2 ** attempt) + random.random()
            print(f"Rate limited. Waiting {wait:.1f}s...")
            time.sleep(wait)

• Use tenacity in production: the tenacity library provides battle-tested retry decorators with jitter and max-wait configuration.
• Tier progression: Tier 1 unlocks after a $5 payment; Tier 2 after $50 cumulative spend at 7 days; higher tiers unlock higher RPM/TPM limits.
• Check billing limit: platform.openai.com/settings/billing → Usage limits — raise or remove the monthly cap.

500 / 503 Server Errors

These are OpenAI infrastructure errors — not caused by your code. Check prismix.dev/service/openai or status.openai.com for active incidents. Always implement retry for 5xx errors:

• 500 Internal Server Error: transient OpenAI-side failure — retry with backoff, usually resolves within seconds.
• 503 Service Unavailable: OpenAI is under heavy load or in maintenance — check status page and retry.
• 502 Bad Gateway: request timed out upstream — reduce prompt length, lower max_tokens, or switch to a faster model like gpt-4o-mini.

Context length exceeded

The error message includes the exact limit: "maximum context length is X tokens but you requested Y tokens". Solutions:

• Switch to gpt-4o: supports 128K tokens context — large enough for most documents and long conversations.
• Truncate conversation history: keep only the last N messages in the messages array instead of the full history.

Count tokens before sending with tiktoken:

import tiktoken
enc = tiktoken.encoding_for_model("gpt-4o")
tokens = len(enc.encode(your_text))
print(f"Token count: {tokens}")

Streaming issues

Use the context manager form and iterate chunks — do not call response.content on a streaming response:

# Python streaming
with client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Tell me a story"}],
    stream=True,
) as stream:
    for chunk in stream:
        if chunk.choices[0].delta.content:
            print(chunk.choices[0].delta.content, end="")

• Do not access response.content on a stream: iterate chunk.choices[0].delta.content inside the loop — accessing the full response at once raises an error on a streaming object.
• Usage stats in streaming: token counts (prompt_tokens, completion_tokens) are only included in the final chunk when you pass stream_options={"include_usage": true}.