Hugging Face's `transformers` library has released version 5.10.1, addressing an issue with a previous corrupted release. This update introduces several new models, including Gemma 4 12B Unified for multimodal tasks, Sapiens2 for high-resolution human-centric vision, DeepSeek-OCR-2 for advanced document understanding, and Mellum, a code-focused MoE model from JetBrains. The release also includes significant bug fixes for model parallelism across various architectures. AI
IMPACT Expands the Hugging Face ecosystem with new multimodal, vision, OCR, and code models, offering developers more options for diverse AI tasks.
RANK_REASON This is a library release with multiple new model additions and bug fixes, fitting the 'research' bucket for OSS model releases.
Read on Transformers — Releases →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
