Abu Dhabi-based research institute IFM on Thursday launched six artificial intelligence models under the K2 Horizon package, providing full model weights, training datasets, source code, methodologies, and intermediate checkpoints. The move aims to enable researchers to reconstruct and verify the development process, distinguishing the release from conventional open-weight approaches that offer limited construction details.
The initiative was announced by IFM founder Eric Xing, who emphasized the institute’s goal of setting a benchmark for transparency in AI model releases. Unlike some Chinese developers that distribute models without full disclosure of underlying data and techniques, or companies such as OpenAI and Anthropic that restrict access to their models and proprietary information, IFM’s release provides complete technical documentation.
IFM stated the launch seeks to demonstrate that openness in AI development does not preclude competitive performance. The institute added that the initiative aligns with broader efforts by the United Arab Emirates to establish itself as a global center for artificial intelligence innovation.
