GPT-5.2 context limits

Real input and output token limits for models, which can differ from the model's documented context window.

A model's advertised context window is the total token budget shared between input (prompt) and output. Some models cap the input and output portions below that total, so the maximum prompt you can send is smaller than the full context window.

openai/gpt-5.2

openai/gpt-5.2 accepts at most 272,000 input (prompt) tokens and emits up to 128,000 output tokens, even though the model's full context window is documented as 400,000 tokens. Requests that exceed 272,000 prompt tokens are rejected with an exceeds the context window error.

This limit applies consistently across providers:

• Azure: served with a 400,000 total context window, split as 272,000 input + 128,000 output. See Azure's model reference.
• OpenAI direct: additional validation testing shows the same behavior — 272,000 input + 128,000 output. See OpenAI's GPT-5 for developers announcement.