Apertus 1.5 Release Notes

Apertus 1.5 extends the previous 8B and 70B models with significant new capabilities, including native understanding of images and audio. The update introduces a "thinking mode" for improved reasoning, expands the context window fourfold to 262,144 tokens, and enhances instruction-following and tool use. Training involved an additional 4 trillion tokens for the 8B model and 2 trillion for the 70B model, incorporating multimodal data through an early-fusion tokenizer approach. The entire project, including model weights, data sources, and training details, is released openly under an Apache 2.0 license.