True open-source LLM Apertus 1.5 released (www.apertus-ai.org)

The release of Apertus 1.5 marks a significant advancement in open-source large language models, enhancing its predecessors with notable features such as native image understanding, a thinking mode for improved reasoning, and a dramatically increased context window of 262,144 tokens—four times larger than Apertus 1.0. This new model integrates both text and multimodal data, processing images, diagrams, and spoken language, which opens up new avenues for applications in areas like technical documentation and complex problem-solving. Apertus 1.5 is built on an extensive pretraining process, incorporating 4 trillion tokens of diverse data for the 8B model and 2 trillion for the 70B model. With a focus on transparency and community engagement, Apertus maintains its commitment to open values by providing full access to the model weights, training data, and detailed technical specifications. This release not only enhances the functionality and usability of the models but also fosters collaboration within the AI/ML community, encouraging developers and enterprises to leverage and contribute to the Apertus ecosystem.