Transformers Release 5.18.0
New Model additions Nemotron 3 Diarization Nemotron 3 Diarization is an open-weight streaming speaker diarization model designed to determine "who spoke when" in real-world audio. It supports both streaming and offline inference, handles up to eight speakers, and orders speaker outputs by each speaker's first arrival in the input audio.
Also shipped Sep 30, 2026Hugging Face in September 2026 →
Amazon Aurora PostgreSQL now supports direct querying of Apache Iceberg and Parquet data in your data lake
Amazon Aurora PostgreSQL now supports direct queries of data stored in Apache Iceberg and Parquet formats within data lakes. This eliminates the need for ETL pipelines when combining transactional data with historical records.
Place pipeline processors anywhere you need them with Pipeline Control
Pipeline Control pipelines used to ship with a static, fixed structure , one Sample Rate node, one Filter node, one Transform node, in a set order. Now you can add and delete Sample Rate, Filter, and Transform processors at any point in the pipeline, in any order, as many times as you need.
The updated version of Pathpoint is now available in Public Preview
The updated version of Pathpoint is now available in Public Preview. When something breaks in a complex system, sometimes the hardest part isn't fixing it, but figuring out how it may impact the business and what part of the customer journey to look at first.
The new Astro experience is now live for all Organizations
The new Astro experience is now turned on for all Organizations and is out of Preview. It replaces the previous Astro experience, so there is no longer an opt-in or an option to switch back. The change doesn't affect your pipelines or configuration. Astro's navigation gives you a single view across your entire Organization instead of a single Workspace.
The Astro IDE is generally available
The Astro IDE is now generally available and is no longer a Preview feature. The Astro IDE is a browser-based workspace for Apache Airflow where you write Dag code, validate and test changes, and deploy or push your work without setting up a local development environment.
Containers - New scheduling policy for Containers to configure image and instance from Durable Objects
Containers now support the durable_object scheduling policy in public beta. This policy lets a Durable Object select the image and instance size for a Container at runtime instead of using one centrally managed configuration for the application.
More from TransformersEvery Transformers release →
Transformers Release v5.19.0
Release v5.19.0 New Model additions EmbeddingGemma2 EmbeddingGemma 2 is a multimodal embedding model from Google built on the Gemma 4 architecture. It encodes text, images, audio, and video, individually or combined in one input, into a shared 768-dimensional vector space for cross-modal retrieval, semantic similarity, clustering, and classification.
Transformers Release v5.20.0
Release v5.20.0 New Model additions Muse Spark Open Muse Spark Open is a multimodal mixture-of-experts model from Meta that takes text, images, audio, and video as input and generates text. It has 816B total parameters (42B active per token) and a context window of up to 1,048,576 tokens shared across modalities.
Transformers Release v5.20.1
Release v5.20.1 Special surprise including the new model Beam from Reflection AI :partying_face: Beam is a token-efficient, frontier-scale model focused on coding and agentic use cases. It is built for teams that want to run a powerful model on their own hardware, infrastructure, or cloud, and customize it to their specific needs.
Transformers Patch release v5.20.2
Patch release: v5.20.2 This patch notably solves an issue where some inputs could loose their original type inside prepare_inputs_for_generation in generate, leading to problems with multimodal models. It also fixes some other issues listed on the below commits.
Transformers Release 5.17.0
Release v5.17.0 New Model additions HYV4 Hy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per token. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every token to 8 of them. The context window is 1M tokens.
ThinkingBox-Training released
ThinkingBox-Training is now available to post-train agents using executable database state as a reward. The framework uses offline policy-relative difficulty to select training tasks and improve agent dependability.
Transformers Release: v5.21.0
Transformers v5.21.0 introduces the MiniCPM-OCR document-understanding model and the Flint mixture-of-experts language model. These additions expand the library's support for specialized vision and language tasks.
Transformers 5.22.0: Mistral Large 4 (Le Chonk)
Transformers v5.22.0 adds support for Mistral Large 4, a 1-trillion parameter multimodal Mixture-of-Experts model featuring a 1-million token context window.