Intel / repo

openai-python v3.45.0

Source: github.com
Author: openai-sdks[bot]Top Viber
Date: Published Nov 4, 2026

Context

Version 3.45.0 of OpenAI's official Python library adds support for two API features: 12-hour caching on the Responses API, with usage reporting for cache writes, and settings for a project's service tier (release notes). For Python developers, that means these two options are now available through the SDK itself. The only other change is a documentation note on which files are generated.

lets the API reuse work it already did on a repeated prompt prefix, such as the same instructions and tool definitions sent on every request, and bill those at a lower cached rate. A longer cache lifetime matters for agents whose requests arrive hours apart; OpenAI's developer account announced selectable 12-hour caching in October, and this SDK release catches the Python client up on the Responses side. The changelog names the features but not their field names, defaults or pricing, so the pull requests linked from the release notes and OpenAI's API documentation are the place to confirm how to use them.

Terms in this piece · Glossary

prompt caching — Reusing the model's processed form of a repeated prompt prefix so subsequent calls skip re-reading it, cutting cost and latency substantially.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.

Read the source github.com