Abu Dhabi-based institute releases open-source artificial intelligence models along with training data
Challenging the trend toward closed-source models, the IFM institute has shared six new artificial intelligence models including weights, training data, code, and methodologies.
The Abu Dhabi-based research institute IFM has made six new artificial intelligence models publicly available, including model weights, training data, source codes, and development methodologies, in defiance of the secrecy trend in the AI sector.
A True Open-Source Approach
The K2 Horizon release includes model weights, training data, codes, methodologies, and intermediate checkpoints, allowing researchers to retrace the development process and fully reproduce the results.
This move goes far beyond the open-weight approach of some Chinese developers who only offer download capabilities while providing limited information about their development processes.
Stance Against Global Trends
These shared models stand in direct contrast to closed development practices by companies such as OpenAI and Anthropic, which neither release their models for free downloading nor disclose the data and techniques behind them.
IFM founder Eric Xing stated that with this release, they aim to establish a benchmark for what truly open model sharing can look like.
Balance Between Transparency and Performance
This launch aims to demonstrate to policymakers, regulators, and public interest advocates that transparency and competitive performance are not mutually exclusive.
This initiative also forms part of the United Arab Emirates' strategic efforts to become a global hub for artificial intelligence.
Models Tailored to Diverse Needs
The released model family covers a broad spectrum, ranging from small systems designed for resource-constrained devices like smartwatches to massive systems with 375 billion parameters developed for enterprise use.