AI news for builders and product teamsUpdated Nov 17, 2026, 13:01 UTC
Hugging Face Blog
First-party releases and research from Hugging Face Blog. Headlines and excerpts link to the original articles.
Latest stories
Newest first
Sentence Transformers v6.0 introduces a fourth model type, MultiVectorEncoder, for ColBERT-style late interaction retrieval, along with a full training approach. A Hugging Face blog post documents finetuning multi-vector models, including a medical retrieval model trained in 14.5 hours on one RTX 3090.

IBM's Granite Team published a technical walkthrough of Granite 4.2, its first family of dense, decoder-only reasoning LLMs, released in 3B, 8B, and 30B sizes under the Apache 2.0 license. The models are post-trained from Granite-4.1 base models via SFT and a multi-stage RL pipeline.

Multiverse Computing researchers introduced Quantization-Aware Healing (QAH), a recipe that distills from the original pre-compression model rather than the recovered checkpoint. Applied to a GPT-OSS 120B compressed to 60B and quantized to MXFP4, the resulting model beat its bfloat16 source on 7 of 9 benchmarks, but an author acknowledged the headline table compares checkpoints with unequal training and lacks a control.

Gradio introduced gr.Workflow, a tool that turns AI pipelines into a drag-and-drop graph of typed nodes where each step is runnable and every intermediate result is visible. Each workflow also serves as a REST API and can be deployed to Hugging Face Spaces with one command.

Hugging Face research introduces three tests to quantify benchmark optimization in speech recognition, finding that several top-scoring open-source ASR models reproduced benchmark reference transcripts even when the audio contradicted them, words were silenced, or multiple written forms were equally supported.

Hugging Face detailed how Papers with Code uses Hugging Face Jobs, Storage Buckets, and Inference Endpoints to build a hybrid search system combining PostgreSQL full-text search with pgvector semantic search, covering more than 110,000 papers from arXiv and Daily Papers.

Liquid AI released DSpark draft model checkpoints for three LFM2.5 models, adding speculative decoding that it reports delivers up to 3.18x throughput improvement on GPU and up to 2.87x on-device without changing output quality.

A Hugging Face blog post reports that ALTK-Evolve, which distills guidelines from an agent's own past trajectories and injects them at inference time without weight updates, improves task completion only when the amount of memory is calibrated to the model. Across eight models, strong models benefited from the full guideline set while weaker models did best with curated retrieval.

Sentence Transformers v6.0 adds a fourth model type, MultiVectorEncoder, for ColBERT-style late interaction retrieval, per a Hugging Face blog post. The post says PyLate and Stanford-NLP ColBERT checkpoints load directly, colpali-engine visual document retrieval models can be used after per-repository configuration, and it describes MaxSim scoring plus index-size tradeoffs.

Dharma AI benchmarked a constraint-aware GPU allocator against a FIFO scheduler across seven scenarios, reporting GPU utilization gains of up to 33 percentage points and priority-weighted output increases of up to 105% on identical hardware and workloads.

Hugging Face's summer 2026 open-models report observes Hub repository growth from January to August 2026, covering download and like distributions, Chinese and US model release sizes, licensing, Qwen derivative counts, small-model usage, local inference formats, and agent traffic. It also notes a July agent intrusion and that analysis was completed on a quantized open model.

Hugging Face published a walkthrough of a streaming data loop using Strands Robots, LeRobot, and Hugging Face Storage Buckets, covering recording robot demonstrations, training by streaming datasets directly from the Hub, and deploying checkpoints back to hardware.