LiteLLM v1.101.0 teaches its complexity router to escalate oversized prompts and route image requests by modality

GitHub·medium signalXBlueskyLinkedInCopy link

Released 2026-09-15, the auto-router work is the substantive part of a 60-item changelog: oversized prompts now escalate to a tier that fits before dispatch rather than failing at the provider, a classification_mode setting skips the classifier on continuation turns, and opt-in modality-based capability routing sends image requests to models that can handle them. shadow_eval gains the ability to compare several auto-routers on one job's sampled traffic and to target teams and users so JWT-auth traffic can be evaluated. The release also adds OIDC workload identity federation for OpenAI, a /v1/responses/input_tokens counting endpoint, and pricing for zai-org/GLM-5.3 and GLM-5.3-Flash on Friendli.