Reddit
K2 Horizon 7B lands between Qwen 3.6 27B and 35B-A3B on Artificial Analysis, and is actually 9B
IFM's K2 Horizon lineup showed up on the Artificial Analysis Intelligence Index, and the 7B slots between Qwen 3.6 27B and Qwen 3.6 35B-A3B, which would be remarkable for the size. Two corrections from the thread and the model card: the Hugging Face config reports roughly 8.999B parameters in BF16 for a dense model, and the README states 18.0 GiB of KV cache at 128K tokens, which the top comment argues disqualifies it from the 'GPU poor' framing the post used. The repo is Apache 2.0, created 2026-09-01, with 7,447 downloads and 152 likes.
Source
↳ Follow the thread