IFM Introduces K2 Horizon Open AI Model Fleet
IFM has announced K2 Horizon, a fleet of six open AI models built for reasoning, coding, agentic workflows, edge devices, and enterprise deployment. The suite features models of various sizes, including the flagship K2 Horizon 375B A23B designed to deliver frontier-level intelligence. Releasing open models capable of competing at the frontier gives researchers and developers powerful alternatives to proprietary commercial APIs. This enables organizations to run sophisticated reasoning and coding systems on their own infrastructure while maintaining control over privacy and security. The lineup covers deployment needs from small edge models like K2-Horizon-7B up to large-scale architectures like K2 Horizon 375B A23B, which scored 47 on the Artificial Analysis Intelligence Index. Accessing certain model weights on Hugging Face requires agreeing to specified usage conditions and sharing contact details.
## BACKGROUND
Open-weights AI models allow developers to run, adapt, and deploy language models directly on local hardware without depending on third-party cloud services. As proprietary frontier models advance rapidly in complex reasoning and software development tasks, open-source initiatives aim to close the capability gap for the broader tech community.