What happened
An entry in the swiss-ai organisation's Hugging Face blog space, dated October 15, 2026, unveiled Apertus 1.5 as a refresh of the 8B and 70B Apertus 1.0 models. The release adds the ability to interpret images and audio, an optional thinking mode, and a 262,144-token context window, which is four times the size offered by Apertus 1.0. Weights for the refreshed 8B and 70B models became available on Hugging Face on July 24, 2026, published under the Apache 2.0 licence. Swisscom, Infomaniak and Phoeniqs each said they would offer API access to developers.
Why it matters
People can now pair text with images and audio when working with these models, and can turn on a thinking mode so the models reason through an input before responding. The models produce only text as output.
Read the original at Hugging Face