v1.99.4 - End-User Budget Resets and GPT-6 Request Mapping
Deploy this version​
- Docker
- Pip
docker run \
-e LITELLM_MASTER_KEY=sk-<paste-a-long-random-key> \
-e DATABASE_URL=postgresql://<user>:<password>@<host>:5432/<dbname> \
-e STORE_MODEL_IN_DB=True \
-p 4000:4000 \
docker.litellm.ai/berriai/litellm:1.99.4
pip install litellm==1.99.4
This release is published as ghcr.io/berriai/litellm:v1.99.4. See the GitHub release and the full releases page
v1.99.4 is a patch release on top of v1.99.3. It fixes end-user budget resets and maps GPT-6 model names to the GPT-5 request family. There are no new database migrations or breaking changes. The v1.99.4 tag points at b6f084f
End-user budget resets​
A shared budget with more than about 32,700 end users never reset, because the reset job listed every end user by id in one statement and Postgres rejected it, so those end users stayed blocked. The job now resets end users by their budget link. A reset also zeroes the cached end-user spend counter in memory and Redis, so requests on every replica stop getting a 429 once the window rolls over instead of after the cache expires
GPT-6 model names use the GPT-5 request family​
OpenAI and Azure configs now treat gpt-6 model names like gpt-5, so they get the same request parameter handling. The lockfile also refreshes anyio, gitpython and soupsieve
What's Changed​
- fix: treat gpt-6 names as the gpt-5 request family in OpenAI and Azure configs - PR #39631
- fix(proxy): invalidate end-user spend counter and cache on budget reset - PR #39729
- fix(reset_budget_job): reset end users by budget link, not by user id - PR #40639
Full Changelog​
https://github.com/BerriAI/litellm/compare/v1.99.3...v1.99.4
