How to Fix 'Context Length Exceeded' Token Limit Errors

A context_length_exceeded error means your prompt plus the requested output is larger than the model's context window. Every model has a fixed ceiling measured in tokens, not characters. The fix is to measure how many tokens you are sending, trim what you can, and split anything too big into chunks.

• Your model's context window size, from the provider docs
• A token counting library such as tiktoken
• The long input that triggered the error

Step 1: Read what the error tells you

The message usually states the model maximum and how many tokens you tried to send. That gap is exactly how much you need to cut, plus headroom for the response.

Step 2: Count tokens before you send

Stop guessing. Count tokens locally so you can reject or trim oversized input before it ever hits the API. A rough rule is about four characters per token in English, but a real counter is exact.

import tiktoken

enc = tiktoken.encoding_for_model("gpt-4o")
text = open("doc.txt").read()
n = len(enc.encode(text))
print(f"{n} tokens")
if n > 120000:
    print("Too big, must chunk")

Step 3: Trim what you do not need

Before chunking, remove fat. Drop old chat history, strip boilerplate, and reserve room for the answer by leaving the max_tokens output budget out of your input total.

Step 4: Chunk long documents

If the input is genuinely larger than the window, split it into overlapping chunks, process each, then combine the results. Overlap a few hundred tokens between chunks so you do not cut a sentence or idea in half.

def chunk(tokens, size=20000, overlap=500):
    out = []
    start = 0
    while start < len(tokens):
        out.append(tokens[start:start + size])
        start += size - overlap
    return out

Result

After counting tokens and splitting a 200-page PDF into overlapping 20k-token chunks, a summarizer that always failed now processes the whole document and merges the chunk summaries into one clean output.