NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #97 most downloaded on PyPI
The official Python library for the openai API
Latest stable
3.55.1
Last release today
18 Nov 2026
Ships on a steady schedule
a new release about every 9 days
Nearly every release is documented
notes for 60 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
7 years old
476 releases · first in 2020
Drag across the ruler to compare, or tap a release to open it.
internal: remove some duplicated imports
internal: add support for TypeAliasType
Your coding agent can read these notes before it upgrades. Set up the MCP server →
azure: handle trailing slash in azure_endpoint
internal: bump pydantic dependency
## 1.57.0 (2024-12-05) Full Changelog: v1.56.2...v1.57.0 ### Features * api: updates (#1924) (82ba614) ### Chores * bump openapi url
## 1.56.2 (2024-12-04) Full Changelog: v1.56.1...v1.56.2 ### Chores * make the Omit type public
cli: remove usage of httpx proxies
Full Changelog: v1.56.0...v1.56.1
client: make ChatCompletionStreamState public
client: compat with new httpx 0.28.0 release
internal: exclude mypy from running on tests
pydantic-v1: avoid runtime error for assistants streaming
Full Changelog: v1.55.0...v1.55.1
cached-property dep (#1867) (df5fac1)cached-property dep (#1891) (feebaae)api: add gpt-4o-2024-11-20 model
asyncify: avoid hanging process under certain conditions
don't use dicts as iterables in transform
## 1.54.3 (2024-11-06) Full Changelog: v1.54.2...v1.54.3 ### Bug Fixes * logs: redact sensitive headers
tests: adjust retry timeout values
add new prediction param to all methods
Full Changelog: v1.54.0...v1.54.1
api: add support for predicted outputs
don't use dicts as iterables in transform
api: add new, expressive voices for Realtime and Audio in Chat Completions
Full Changelog: v1.52.2...v1.53.0
## 1.52.2 (2024-10-23) Full Changelog: v1.52.1...v1.52.2 ### Chores * internal: update spec version
client/async: correctly retry in all cases
api: add gpt-4o-audio-preview model for chat completions
## 1.51.2 (2024-10-08) Full Changelog: v1.51.1...v1.51.2 ### Chores * add repr to PageInfo class
client: avoid OverflowError with very large retry counts
api: support storing chat completions, enabling evals and model distillation in the dashboard
Full Changelog: v1.50.2...v1.51.0
audio: correct types for transcriptions / translations
helpers: fix chat completion anchor
structured outputs: add support for accessing raw responses
internal: update test snapshots
client: allow overriding retry count header
pydantic v1: avoid warnings error
Full Changelog: v1.47.0...v1.47.1
client: send retry count header
Full Changelog: v1.46.1...v1.47.0
client: handle domains with underscores
client: add ._request_id property to object responses
internal: bump pyright / mypy version
errors: include completion in LengthFinishReasonError
add docstrings to raw response properties
vector store: improve chunking strategy type names
pyproject.toml formatting changes
api: add file search result details to run steps
parsing: add support for pydantic dataclasses
json schema: remove None defaults
client: add uploads.upload_file helper
Full Changelog: v1.40.8...v1.41.0
## 1.40.8 (2024-08-15) Full Changelog: v1.40.7...v1.40.8 ### Chores * types: define FilePurpose enum
cli/migrate: change grit binaries download source
examples: minor formatting changes
helpers: make async client usage more clear (34e1edf), closes #1639
json schema: unravel $refs alongside additional keys
ci: bump prism mock server version
client: raise helpful error message for response_format misuse
internal: update OpenAPI spec url
api: add structured outputs support
client: add retries_taken to raw response class
Full Changelog: v1.38.0...v1.39.0
extract out ImageModel, AudioModel, SpeechModel
internal: add type construction helper
## 1.37.1 (2024-07-25) Full Changelog: v1.37.0...v1.37.1 ### Chores * tests: update prism version
cli/audio: handle non-json response format
types: add gpt-4o-mini to more assistants methods
Full Changelog: v1.36.0...v1.36.1
api: add new gpt-4o-mini models
One column per quarter.