Apertus 1.5 adds image and audio input with expanded context

Project Apertus has released Apertus 1.5 with open weights, training data and training details, adding image and audio input to its 8B and 70B models. Its release notes describe a 262,144-token context window, four times that of Apertus 1.0, and a switchable reasoning mode. The models process images and audio as tokens alongside text and generate only text. Intermediate training checkpoints and curated data mixes are available on Hugging Face. Proton Lumo and Infomaniak Euria have integrated the models. Spoken-language processing remains experimental, and the developers say performance still trails frontier open-weight models.