GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
ATinyGreenCell
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
ATinyGreenCell
Model Tree
Qwen/Qwen3.5-9B
Input Modalities
Output Modalities
Supported Functionality
v7 is v4 trained for ONE epoch instead of three (everything else identical): 9,136 rows = the 7,238 render-verified examples of v2 + 269 rejection-sampled generations (x2) + 680 Claude Opus distilled examples on 17 part families that are not Prompt-to-Bench tasks (x2; every description shares at most 36% of its tokens with any benchmark prompt). LoRA rank 32, dropout 0.1, language layers only. It tests whether lighter training keeps more of the base model. Details: https://github.com/ATinyGreenCell/prompt-to-bench (§6.4).