AI BriefingSwiss-AI releases multimodal Apertus-v1.5
·2026.07.24 15:29
Key point
Swiss-AI has released Apertus-v1.5, an open-source model equipped with multimodal capabilities and a thinking mode.
Details
Swiss-AI has released the Apertus-v1.5 model family (8B and 70B parameters). The models aim for full openness (Open Weights + Open Data) and offer a transparent training process.
Key features include:
- Multimodal support: In addition to text, the models can process image and audio inputs.
- Thinking Mode: A mode was introduced that allows the model to go through its own reasoning process before generating a response, improving performance on complex tasks.
- Large-scale context: Supports a context length of up to 262,144 tokens.
- Multilingual and tool use: Supports a wide range of languages, with improved instruction-following and tool-use capabilities achieved through enhanced post-training.
This version is the result of Continued Pretraining on top of Apertus 1.0, with an additional multimodal mix of 4T tokens added for the 8B model and 2T tokens for the 70B model.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
LLaDA2.X: A Large-Scale Diffusion Language Model SeriesHugging Face CEO details lessons from first public agent cyberattack, advocates for open-source defense
NASA-IBM Releases Open-Source Multimodal Geospatial AI Model 'Lunar Foundation' for Lunar Exploration