LiteLLM v1.110.0 Adds Parallel Decisions and Provider Updates
LiteLLM v1.110.0 introduces parallel AI web-grounded decisions, custom provider video endpoints, provider compatibility fixes, model deprecation tracking, and new UI improvements.
What changed?
LiteLLM v1.110.0 brings multiple enhancements and fixes, including support for parallel AI web-grounded decisions, a new /v1/videos endpoint for custom video providers, a capability flag for Gemini's long decode continuation, and updated compatibility for several AI providers. The release also introduces improved tracking of model deprecations and replacements, new information in the model catalog, and user interface updates, such as displaying usernames instead of IDs.
- Parallel AI web-grounded decisions are now available.
- Custom provider video endpoints are now supported via /v1/videos.
- The Gemini provider returns a 400 error for unsupported chat audio formats rather than failing silently.
- The model catalog now includes deprecation tracking, successor models, deprecation sources, and a FedRAMP uplift multiplier.
- Anthropic, Vertex AI, and Azure AI header allowlists are now synchronized for better compatibility.
- The UI shows user names instead of raw IDs on detail pages.
- Prometheus attribution is improved for deployment failures.
Why does it matter to an everyday developer?
The update improves reliability and compatibility for developers building applications with LiteLLM across multiple AI providers. The new parallel web-grounded decisions enable faster and potentially more accurate decision-making by leveraging AI on multiple web sources. Improved error handling in the Gemini provider results in clearer feedback when unsupported formats are used. Model deprecation tracking and successor recommendations help teams manage model lifecycle and plan migrations. Finally, UI changes and new APIs make it easier to integrate, monitor, and manage models and deployments.
