GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
TreezzZ
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
TreezzZ
Model Tree
Qwen/Qwen3-4B
Input Modalities
Output Modalities
Supported Functionality
WebShop success (%) on the 500 canonical goals. Mean ± std over three decodes (seeds 42, 1001, 1002) of the final checkpoint.
| Student alone | Vanilla OPD | R2S (this model) |
|---|---|---|
| 15.5 ± 1.2 | 24.7 ± 1.1 | 28.5 ± 0.6 |
python
from transformers import AutoModelForCausalLM, AutoTokenizerrepo = "TreezzZ/r2s-webshop-qwen3-4b-student"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo, dtype="bfloat16")The checkpoint was trained with the environment prompts, tool schemas and rollout settings of the paper's code release; the chat template is the base model's.
Trained with the paper's code release, Tree-Shu-Zhao/miles-agent at tag r2s (its README reproduces the main table; docs/r2s.md holds the details). The weights were exported from the training checkpoint and verified against the base model's tensor names and shapes.