# Hugging Face Transformers v5.15.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v5.15.0
- **Released:** Aug 10, 2026
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v5.15.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v5.15.0. Source highlight: Muse Glimmer, released today, is Meta’s new multimodal model, especially designed for agentic use cases. Distilled from Muse to 30B parameters, and released under the Apache 2.0 license, it can be deployed to local setups for privacy-aware applications such as coding, document analysis, personal ass

## Breaking Changes & Deprecations
- Breaking changes Kernels are now opt-in rather than mandatory for linear attention models (Mamba, GDN, Conv-only, etc.), so users who relied on automatic kernel selection must explicitly enable kernels to maintain previous behavior.
- Fix CUDA Graph breaking host to device copy from scalar tensor allocation (#47547) by @hmellor in [#47547]

## New Features
- Add Granite-swa and Granitemoe-swa model support (#47179) by @daviswer in #47179
- Add Granite-swa and Granitemoe-swa model support (#47179) by @daviswer in #47179
- Add AXK2 from SKT (#47528) by @vasqu in #47528
- Add Cosmos3 Edge model support (#47181) by @atharvajoshi10 in #47181
- add paged attention tests support for XPU (#47163) by @kaixuanliu in [#47163]
- Align image feature dtype in kosmos2 and kosmos2_5 embedding merge (#47691) by @<NOT FOUND> in [#47691]

## Bug Fixes
- [Fix] Fix multi-head latent attention (MLA) (#47761) by @remi-or in [#47761]
- Simplify function dispatch for linear attention (#47450) by @Cyrilvallez in [#47450]
- Fix `BlockMask` crash in CTRL flex-attention generation (#46854) by @jiqing-feng in [#46854]
- Fix sdpa prefill with position_bias (#47359) by @Cyrilvallez in [#47359]
- Fix vision position-embedding init width fallback in Phi4Multimodal (#47509) by @<NOT FOUND> in [#47509]
- Fix Hunyuan-VL PIL image resize parity with reference preprocessing (#47233) by @IMvision12 in [#47233]

## Upgrade Command
```bash
python -m pip install 'transformers==5.15.0'
```