Instructions to use morganstanley/qqWen-72B-RL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use morganstanley/qqWen-72B-RL with Transformers:

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

pipe = pipeline("text-generation", model="morganstanley/qqWen-72B-RL")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("morganstanley/qqWen-72B-RL")
model = AutoModelForCausalLM.from_pretrained("morganstanley/qqWen-72B-RL")
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 morganstanley/qqWen-72B-RL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "morganstanley/qqWen-72B-RL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "morganstanley/qqWen-72B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/morganstanley/qqWen-72B-RL

SGLang

How to use morganstanley/qqWen-72B-RL 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 "morganstanley/qqWen-72B-RL" \
    --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": "morganstanley/qqWen-72B-RL",
		"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 "morganstanley/qqWen-72B-RL" \
        --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": "morganstanley/qqWen-72B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use morganstanley/qqWen-72B-RL with Docker Model Runner:
```
docker model run hf.co/morganstanley/qqWen-72B-RL
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

qqWen-72B-RL: Reasoning-Enhanced Q Programming Language Model

Model Overview

qqWen-72B-RL is a 72-billion parameter language model specifically designed for advanced reasoning and code generation in the Q programming language. Built upon the robust Qwen 2.5 architecture, this model has undergone a comprehensive two-stage training process: pretraining, and reinforcement learning (RL) for the Q programming language. qqWen-72B-RL is a reasoning model.

This model is special/different than the previous models in that the pretraining accuracy is high enough that we are able to skip the SFT phase and directly go to RL. With the hope being this will bias the model less towards pythonic Q.

Associated Technical Report: Report

🔤 About Q Programming Language

Q is a high-performance, vector-oriented programming language developed by Kx Systems, primarily used in:

Financial Markets: High-frequency trading, risk management, and market data analysis
Time-Series Analytics: Real-time processing of large-scale temporal data
Data Science: Efficient manipulation of large datasets with concise syntax
Quantitative Research: Mathematical modeling and statistical analysis

Key Q Language Features:

Vector Operations: Built-in support for element-wise operations on arrays
Functional Programming: First-class functions and powerful combinators
Memory Efficiency: Optimized for handling large datasets in minimal memory
Speed: Exceptional performance for numerical computations
Concise Syntax: Expressive code that can accomplish complex tasks in few lines

📝 Citation

If you use this model in your research or applications, please cite our technical report.

@misc{hogan2025technicalreportfullstackfinetuning,
      title={Technical Report: Full-Stack Fine-Tuning for the Q Programming Language}, 
      author={Brendan R. Hogan and Will Brown and Adel Boyarsky and Anderson Schneider and Yuriy Nevmyvaka},
      year={2025},
      eprint={2508.06813},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2508.06813}, 
}