GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Hoyant-Su
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
Hoyant-Su
Model Tree
Qwen/Qwen3-14B
Input Modalities
Output Modalities
Supported Functionality
bash
hf download Hoyant-Su/A3-Qwen3-14B --local-dir A3-Qwen3-14BThe exported model was validated with Transformers 4.51.1 and PyTorch 2.6.0.
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizer model_id = "Hoyant-Su/A3-Qwen3-14B"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto",)This loads the policy. End-to-end CLI-agent inference additionally requires the task workspace, action parser, shell execution environment, and observation loop provided in the code repository. The vanilla and sigma_reveal_rd harnesses use the same model weights; the harness determines how the agent interacts with the environment.
Refer to the paper and code for benchmark splits, prompting, harness settings, and evaluation procedures. Small demonstration tasks are illustrative and do not reproduce the full benchmark evaluation. Generated shell commands should be executed in an isolated task sandbox.
Released under the Apache License 2.0. The base model is Qwen/Qwen3-14B.
bibtex
@misc{su2026learningcliagentsstructured, title={Learning CLI Agents with Structured Action Credit under Selective Observation}, author={Haoyang Su and Ying Wen}, year={2026}, eprint={2605.08013}, archivePrefix={arXiv}, primaryClass={cs.AI}, url={https://arxiv.org/abs/2605.08013}}@misc{su2026learningcliagentsstructured, title={Learning CLI Agents with Structured Action Credit under Selective Observation}, author={Haoyang Su and Ying Wen}, year={2026}, eprint={2605.08013}, archivePrefix={arXiv}, primaryClass={cs.AI}, url={https://arxiv.org/abs/2605.08013}}