Instructions to use osunlp/QUEST-30B-SFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use osunlp/QUEST-30B-SFT with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="osunlp/QUEST-30B-SFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("osunlp/QUEST-30B-SFT")
model = AutoModelForCausalLM.from_pretrained("osunlp/QUEST-30B-SFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use osunlp/QUEST-30B-SFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "osunlp/QUEST-30B-SFT"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "osunlp/QUEST-30B-SFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/osunlp/QUEST-30B-SFT

SGLang

How to use osunlp/QUEST-30B-SFT with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "osunlp/QUEST-30B-SFT" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "osunlp/QUEST-30B-SFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "osunlp/QUEST-30B-SFT" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "osunlp/QUEST-30B-SFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use osunlp/QUEST-30B-SFT with Docker Model Runner:
```
docker model run hf.co/osunlp/QUEST-30B-SFT
```

hsaest commited on 14 days ago

Commit

ce9f9a8

verified ·

1 Parent(s): 5686633

README: add benchmark result table from project page

Browse files

Files changed (1) hide show

README.md +20 -6

README.md CHANGED Viewed

@@ -9,7 +9,23 @@ tags:
 # QUEST-30B-SFT
-**30B** **vanilla SFT** dense model (QUEST line).
 ## Quick start
@@ -19,14 +35,12 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 model_id = "osunlp/QUEST-30B-SFT"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    device_map="auto",
-    torch_dtype="auto",
 )
 ```
-Use the chat template and `tokenizer.apply_chat_template(...)` when available. VRAM and dtype requirements depend on model size and MoE vs dense architecture; see `config.json` (`model_type`, `architectures`).
 ## License
-This model is released under the **Apache License 2.0** (`apache-2.0`).

 # QUEST-30B-SFT
+QUEST **30B** vanilla SFT checkpoint (Qwen3-30B-A3B base, dense).
+Quantitative results for QUEST-30B (full MT+SFT+RL pipeline) are shown below for reference. SFT-only scores are not separately reported on the project page.
+## QUEST-30B benchmark results (MT+SFT+RL)
+| Benchmark | Metric | Score |
+| --- | --- | ---: |
+| BrowseComp | avg@3 | 37.0 |
+| Mind2Web 2 | avg@3 | 28.6 |
+| HLE | avg@3 | 24.6 |
+| DeepResearch Bench | avg@3 | 45.3 |
+| BrowseComp-Plus | avg@3 | 48.2 |
+| WideSearch | Item F1 avg@4 | 54.2 |
+| GAIA | avg@3 | 69.0 |
+| LiveResearchBench | avg@3 | 74.1 |
+> Results from the full QUEST-30B (MT+SFT+RL) run, not this SFT-only checkpoint.
 ## Quick start
 model_id = "osunlp/QUEST-30B-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**.