# Hugging Face Transformers v5.20.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v5.20.0
- **Released:** Oct 15, 2026
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v5.20.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v5.20.0. Source highlight: <img width="2400" height="1296" alt="image" src="https://github.com/user-attachments/assets/bf34ea27-0edf-4ded-ba6f-905b04096126" />

## Breaking Changes & Deprecations
- Breaking changes 🚨 General breaking notice that inference with EP has regressions in speed; we are working on fixing this ASAP! 🚨
- 🚨 General breaking notice that inference with EP has regressions in speed; we are working on fixing this ASAP! 🚨

## New Features
- Add Muse Spark Open (#49644) by @guarin in #49644
- [Model] Add native Ling 3.0 Flash support (#49446) by @icavan in #49446
- Add Kolibri 1 (#49463) by @lucamotz in #49463
- Add Unlimited OCR (#46836) by @guarin in #46836
- Add MOSS-Transcribe-Diarize model (#48386) by @itazap in #48386
- Add job to automatically generate Intel CPU Docker image (#49560) by @regisss in [#49560]

## Bug Fixes
- Fix more stuff (#48979) by @zucchini-nlp in #48979
- fix(distributed): keep enable_expert_parallel as EP alone (#49633) by @dacorvo in [#49633]
- Fix a port allocation race in FSDP tests (#49455) by @icavan in [#49455]
- Expert-parallel dispatch hands the experts the received rows, not the all-to-all's pending result (#49421) by @IlyasMoutawwakil in [#49421]
- Fix static-cache generate() with host-side inputs and under TP (#49317) by @dacorvo in [#49317]
- Fix: use torch.accelerator device APIs so TP loading works again on TPU (#49502) by @tengomucho in [#49502]

## Upgrade Command
```bash
python -m pip install 'transformers==5.20.0'
```