README.md

---
library_name: transformers
license: apache-2.0
pipeline_tag: text-generation
tags:
- quest
- text-generation
---

# QUEST-35B-SFT

QUEST **35B-class MoE** SFT-only checkpoint (Qwen3.5-35B-A3B base, `Qwen3_5MoeForConditionalGeneration`). Intermediate stage before mid-training and RL.

## Benchmark results

| Benchmark | Metric | Score |
| --- | --- | ---: |
| BrowseComp | avg@3 | 45.1 |
| Mind2Web 2 | avg@3 | 26.5 |
| HLE | avg@3 | 39.49 |
| DeepResearch Bench | avg@3 | 36.35 |
| BrowseComp-Plus | avg@3 | 57.9 |
| WideSearch | Item F1 avg@4 | 61.1 |
| GAIA | avg@3 | 83.5 |
| LiveResearchBench | avg@3 | 64.69 |

## QUEST Family

| Type | Resources |
| --- | --- |
| 35B checkpoints | [RL](https://huggingface.co/osunlp/QUEST-35B-RL), [MT+SFT](https://huggingface.co/osunlp/QUEST-35B-MT-Plus-SFT), [MT](https://huggingface.co/osunlp/QUEST-35B-MT), [SFT](https://huggingface.co/osunlp/QUEST-35B-SFT) |
| 30B checkpoints | [RL](https://huggingface.co/osunlp/QUEST-30B-RL), [MT+SFT](https://huggingface.co/osunlp/QUEST-30B-MT-Plus-SFT), [SFT](https://huggingface.co/osunlp/QUEST-30B-SFT) |
| Smaller checkpoints | [9B](https://huggingface.co/osunlp/QUEST-9B), [4B](https://huggingface.co/osunlp/QUEST-4B), [2B](https://huggingface.co/osunlp/QUEST-2B) |
| Training data | [RL data](https://huggingface.co/datasets/osunlp/QUEST-RL-Data), [SFT objective data](https://huggingface.co/datasets/osunlp/QUEST-SFT-Data-Objective), [SFT open-ended data](https://huggingface.co/datasets/osunlp/QUEST-SFT-Data-Open-ended), [Mid-training data](https://huggingface.co/datasets/osunlp/QUEST-Mid-Training-Data) |

Model selection note: if you only need to evaluate objective tasks and do not
need open-ended task evaluation, we recommend the MT+SFT checkpoints because
they perform better on reasoning-heavy objective benchmarks. For a more comprehensive evaluation
across both objective and open-ended tasks, we recommend the RL checkpoints.

## Quick start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "osunlp/QUEST-35B-SFT"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id, device_map="auto", torch_dtype="auto",
)
```

Apply the model's chat template with `tokenizer.apply_chat_template(...)` before passing prompts.

## License

Released under the **Apache License 2.0**.

## Citation

If our paper or related resources prove valuable to your research, we kindly ask
for a citation.

```bibtex
@misc{xie2026quest,
  title={QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks},
  author={Xie, Jian and Lin, Tianhe and Wang, Zilu and Ning, Yuting and Yao, Yuekun and Xue, Tianci and Zhang, Zhehao and Li, Zhongyang and Zhang, Kai and Wu, Yufan and Chen, Shijie and Gou, Boyu and Han, Mingzhe and Wang, Yifei and Lee, Vint and Wei, Xinpeng and Wang, Xiangjun and Su, Yu and Sun, Huan},
  journal={arXiv preprint arXiv:2605.24218},
  year={2026}
}
```