from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "gulf-inference/phi-4-mini-enhanced",
    trust_remote_code=True,
    torch_dtype="auto"
)
tokenizer = AutoTokenizer.from_pretrained("gulf-inference/phi-4-mini-enhanced")

Key Improvements

Faster inference: optimized attention
Extended context: up to 8K tokens
Drop-in replacement — same API, better performance

License

MIT