Instructions to use OpenGenerativeAI/Bifrost with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OpenGenerativeAI/Bifrost with Transformers:

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

pipe = pipeline("text-generation", model="OpenGenerativeAI/Bifrost")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("OpenGenerativeAI/Bifrost")
model = AutoModelForCausalLM.from_pretrained("OpenGenerativeAI/Bifrost")
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 OpenGenerativeAI/Bifrost with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/OpenGenerativeAI/Bifrost

SGLang

How to use OpenGenerativeAI/Bifrost 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 "OpenGenerativeAI/Bifrost" \
    --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": "OpenGenerativeAI/Bifrost",
		"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 "OpenGenerativeAI/Bifrost" \
        --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": "OpenGenerativeAI/Bifrost",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use OpenGenerativeAI/Bifrost with Docker Model Runner:
```
docker model run hf.co/OpenGenerativeAI/Bifrost
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Bifröst

Bifröst is an advanced AI model built upon Phi-4 integrated into the Llama architecture, specifically fine-tuned for secure and efficient enterprise-grade code generation. Designed to meet rigorous standards of safety, accuracy, and reliability, Bifröst empowers organizations to streamline software development workflows while prioritizing security and compliance.

Model Details

Model Name: Bifröst
Base Architecture: Phi-4 adapted to Llama
Application: Enterprise Secure Code Generation
Release Date: 07-March-2025

Intended Use

Bifröst is designed explicitly for:

Generating secure, efficient, and high-quality code.
Supporting development tasks within regulated enterprise environments.
Enhancing productivity by automating routine coding tasks without compromising security.

Features

Security-Focused Training: Specialized training regimen emphasizing secure coding practices, vulnerability reduction, and adherence to security standards.
Enterprise-Optimized Performance: Tailored to support various programming languages and enterprise frameworks with robust, context-aware suggestions.
Compliance-Driven Design: Incorporates features to aid in maintaining compliance with industry-specific standards (e.g., GDPR, HIPAA, SOC 2).

Limitations

Bifröst should be used under human supervision to ensure code correctness and security compliance.
Model-generated code should undergo appropriate security and quality assurance checks before deployment.

Ethical Considerations

Users are encouraged to perform regular audits and compliance checks on generated outputs.
Enterprises should implement responsible AI practices to mitigate biases or unintended consequences.

Downloads last month: 21

Safetensors

Model size

15B params

Tensor type

BF16

Model tree for OpenGenerativeAI/Bifrost

Base model

microsoft/phi-4

Finetuned

(95)

this model

Finetunes

1 model

Quantizations

2 models