GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
singhabhishekkk
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
singhabhishekkk
Model Tree
Qwen/Qwen3.5-4B
Input Modalities
Output Modalities
Supported Functionality
| System | Score |
|---|---|
| gpt-5.4-mini, shipped paperless-gpt prompt | 78.33 |
| gpt-5.4-mini, GEPA-optimized | 81.67 |
| Qwen3.5-4B raw | 71.67 |
| Qwen3.5-4B fine-tuned (this adapter) | 80.00 |
The fine-tune beats the frontier model as paperless-gpt ships it (78.33) and lands 1.67 short of the GEPA-optimized teacher.
LoRA r=16, alpha 16, 3 epochs, lr 2e-4, batch 2 x grad-accum 4, Unsloth 4-bit, Colab GPU. Train/eval split: seed 42, 140/60 from 200 sampled rows, identical split across every model this task is fine-tuned on.
Load with PEFT on top of Qwen/Qwen3.5-4B, or serve locally with an adapter-capable runtime. Caveat: evaluated on 60 rows for one field only. Re-validate on your paperless-ngx document types before production use.