Apertus 1.5 adds multimodal input, thinking mode, and 262K-token context to open 8B and 70B models

Apertus 1.5 updates the 8B and 70B Apertus 1.0 models with image and audio understanding, optional thinking mode, improved instruction-following and tool use, and a 262,144-token context window. The Apache 2.0 weights were released on Hugging Face on July 24, 2026, with open data, training details, evaluation records, and a technical report; spoken-language processing remains experimental.

Builders needing auditable models can use Apertus 1.5’s open weights, data, and recipes to inspect training and evaluation claims, but the practical trade-off is capability versus maturity: the article says it still trails frontier open-weight models, generates only text, and treats spoken language as experimental. The 262K context and tokenizer-based multimodality may help long-document and image/audio workflows, but teams still need to validate cost, latency, and task performance on their own workloads.