Apertus 1.5 Release Notes
- Source
- huggingface.co
- Date
Context
Apertus 1.5 extends the fully open Apertus 8B and 70B models with image and audio understanding by turning those inputs into tokens processed in the same sequence as text. The project's release notes describe extending the existing models' vocabulary rather than attaching a separate vision encoder. The models accept multiple kinds of input but still produce text, and spoken-language processing remains experimental; understanding audio does not mean generating speech.
The release also exposes intermediate training checkpoints and data mixtures. Those artifacts give researchers a way to investigate how successive training stages affect behavior, instead of having only the finished model to examine. The notes acknowledge that the models still trail frontier systems overall, so the research value of that access is distinct from a claim to lead every benchmark.
- open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
- pretraining — The first, biggest phase of building a model: training it on enormous amounts of text so it learns language, facts, and reasoning in general.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
Checking sign-in…
Loading comments…