GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
win10
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
win10
Model Tree
win10/K2-nano-v1
Input Modalities
Output Modalities
Supported Functionality
This model was completed using a private model-merging algorithm. Its real-world performance is significantly stronger than the original models, and mathematically, it forms an almost perfect centroid.
This is a checkpoint for cross-architecture merging.
This work demonstrates the successful integration of quantization models with cross-architecture models.
The resulting model quality surpasses that of the same model source under BF16,
showing that quantized model merging can produce results exceeding those of the original BF16-level models,
and also demonstrating that cross-architecture merging without training is a feasible approach.
Special thanks to the authors of the following models:
win10/K2-nano-v1kldzj/gpt-oss-120b-heretic-bf16huihui-ai/Huihui-Nex-N2-mini-abliteratedpoolside/Laguna-XS-2.1Support this research and future model development: