Models & Labs
Apertus 1.5 adds native multimodal and thinking mode
Hugging Face Bloghigh confidence
Why it matters
- →Native multimodal support via early fusion eliminates the need for separate vision encoders in open-weight models.
- →Thinking mode and 262k context window make these models viable for complex reasoning and long-document analysis.
- →Full transparency in training data and evaluation metrics sets a new standard for reproducible open-source AI development.
Apertus has released version 1.5 of its open-weight language models, introducing native support for image and audio inputs alongside text. The update applies to both the 8B and 70B parameter variants, which now feature a 'thinking mode' for improved reasoning and a context window expanded to 262,144 tokens. Training involved continuing pretraining on 4 trillion additional tokens for the smaller model and 2 trillion for the larger one, using an early-fusion approach where visual and audio data are tokenized directly into the text sequence. The release includes full transparency regarding training data, post-training stages, and evaluation metrics aligned with EU AI Act requirements.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More from Hugging Face Blog
More in Models & Labs
Models & Labsmodels
llama.cpp b12029 adds CUDA 13 and Snapdragon support
This release quietly cements llama.cpp as the universal inference runtime by finally supporting NVIDIA's latest CUDA 13 stack alongside AMD's ROCm 10.0. The inclusion of native builds for Linux arm64 with Snapdragon chips, covering CPU, Adreno GPU, and Hexagon NPU, signals a serious push into mobile edge deployment. While Apple Silicon KleidiAI is disabled in this specific build, the broader hardware coverage means developers no longer need to compile from source for most modern accelerators. It’s less about new features and more about removing friction for anyone running local models on bleeding-edge or diverse hardware.
llama.cpp Releases
© The Verge AIModels & Labsother
