Hugging Face released Apertus 1.5 on July 24, 2026, saying it extends the 8B and 70B models with image and audio understanding, an optional thinking mode, a four times longer context window, improved instruction-following, and better tool use. It is the company's first major update since the initial Apertus 1.0 release in 2025.
Hugging Face reported the updated 8B and 70B model weights were released under an Apache 2.0 license, measured on a range of benchmarks and tasks. That compares with the initial Apertus 1.0 models, which were released without multimodal capabilities.
Apertus 1.5 is built on the same architecture as its predecessor and targets applications requiring multimodal understanding, such as document analysis, audio processing, and complex reasoning. Availability begins with full public access, initially for researchers and developers.
"The models can be switched into a thinking mode to reason about the input before answering, which improves performance on reasoning problems," said Oleg Lavrovsky, lead researcher at Hugging Face. The release includes a comprehensive training pipeline and detailed documentation for reproducibility.
The announcement follows the release of the Apertus 1.0 models in 2025, which marked Hugging Face's first major open-weight large language model. Hugging Face said the release reflects its commitment to open-source development and transparent evaluation.
Hugging Face did not say how the models will perform on all tasks, and raised the open question of how the new capabilities will affect real-world applications. The company said it will continue to release updates and improvements in the future.
Source: huggingface
