HuggingFace released Optimum Intel v2.2 and OpenVINO 2026.4 on September 18, 2026, saying it enhances model deployment across multiple AI domains. It is the company's first major update since the Optimum Intel v2.1.0 release on August 6, 2026.
HuggingFace reported the release supports over 10 new models, including Mistral 3, DeepSeek-OCR-2, and Gemma 4, measured on a range of AI tasks. That compares with the previous version, which supported fewer models and tasks.
Optimum Intel v2.2 is built on the OpenVINO platform and targets efficient inference for language, vision, and generative AI applications. Availability begins with the release on September 18, 2026, initially for developers and researchers.
"Optimum Intel 2.2 expands model support across several tasks, letting you export new models to OpenVINO's Intermediate Representation (IR) for efficient inference," said Anastasiia Pnevskaia, HuggingFace's technical lead. The release streamlines the model deployment workflow, from export to inference.
The announcement follows the release of OpenVINO GenAI 2026.4, which integrates new capabilities into task-specific pipelines. HuggingFace said the updates reflect the growing demand for efficient, scalable AI deployment.
HuggingFace did not say when the full list of supported models will be available, and the open question remains about the performance trade-offs of new features. The company said the updates will be further refined in upcoming releases.
Source: huggingface
