IFM Paris has launched K2 Horizon, a suite of six open-source language models ranging in size from 0.9 billion to 375 billion parameters. The models demonstrate top-tier performance across various benchmarks, with the smaller 0.9B, 3.7B, and 7B versions setting new state-of-the-art results for their respective scales. K2 Horizon is notable for its radical openness, providing full access to the training lifecycle, including intermediate checkpoints, data recipes, architecture details, and training code, making it the most comprehensive open release to date for agentic AI development. AI
IMPACT Sets new performance benchmarks across model sizes and provides unprecedented openness for agentic AI research and development.
RANK_REASON Frontier-lab model release with system card and full training lifecycle details. [lever_c_demoted from frontier_release: ic=2 ai=1.0]
Read on Mastodon — mastodon.social →
- K2 Horizon
- 0.9B
- 36B-A4B
- 375B-A23B
- 3.7B
- 7B
- Apache Software License 2.0
- IFM Paris
- Mixture-of-Value-Attention
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
