What changed
Swiss AI published the Apertus 1.5 technical report, documenting continued pretraining, early-fusion image and audio tokenization, optional thinking, tool use and a 262,144-token context for its Apache-2.0 8B and 70B models. The weights themselves were released on 24 July; the new October event is the reproducibility and evaluation record.
Why it matters
The report makes a large multimodal model's data, training stages, values and evaluation choices unusually inspectable. It enables replication and audit work without mislabeling an older checkpoint as a new release, but openness does not establish competitive quality or safe deployment.
What to do
Start with the report's evaluation cards and a pinned 8B checkpoint. Reproduce one long-context, image and audio task, record tokenizer and hardware cost, and compare the July weights with a current baseline before treating the October documentation as a reason to migrate.
Caveat
The authors report their own results, audio support remains experimental, and the models understand images and audio but generate text only. The 8B and 70B weights predate the report by nearly three months, so this record covers new technical evidence rather than new model availability.