GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
armand0e
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
armand0e
Model Tree
Base
Qwen/Qwen3.6-27B
Fine-tuned
this model
Input Modalities
Text
Image
Video
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
Heres Qwen3.6 slightly overfitted on a very small dataset of fable 5 traces XD. I finished this tune yesterday morning, realized it's good for planning but is definitely overcooked on the small amount of data, most code it generates always has a small bug or two.
Had to use aggressive settings for the style transfer here due to the data contraints. Code quality/consitency definitely took a hit, but the model seems to be a better planner now and slightly better at 3D modeling in three.js.
Either way give it a shot, it really does look, talk, and plan like fable... but like all fine-tunes it comes with it's drawbacks.
Wasn't planning on ever releasing this model, but since others are okay with uploading their worse-than-base tunes so frequently I figured someone might enjoy it
Hopefully we can get some more well rounded data from the rest of the community for me to a do a much less aggressive tune.
Reasoning was left untouched
Heres a procedurally generated duck that it made lol (quant: q3_kS)
This qwen3_5 model was trained 2x faster with Unsloth and Huggingface's TRL library.