The K2 Horizon project has released six open-source models, ranging from 0.9 billion to 375 billion parameters. This release is notable for including not only the model weights but also the training code, dataset, logs, and intermediate checkpoints. This comprehensive package aims to make the models fully verifiable and auditable, allowing for independent reproduction and scrutiny of their development process. AI
IMPACT Enables greater transparency and reproducibility in open-source AI development, potentially accelerating research and auditing.
RANK_REASON Release of open-source models with full training data and code, enabling verification and auditing.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →