Phi-4-mini-instruct

Phi-4-mini-instruct (Microsoft)

Modern small instruction-tuned language model with strong reasoning for its size.

• Next-gen Phi. A compact model from Microsoft’s Phi-4 family, trained on high-quality synthetic and curated data, then instruction-tuned for reliable task following and chat.

• Better thinking at small scale. Shows improved reasoning, math, and coding behavior compared to earlier Phi models while staying lightweight.

• Practical context window. Supports longer prompts and multi-turn interactions than older small LLMs, making it usable for real assistants.

• Efficient to deploy. Designed to run on modest GPUs or optimized CPU setups with good throughput.

• Clean setup. Single model, standard tokenizer, no external tools or pipelines required.

Why pick it for Norman AI?

Phi-4-mini-instruct is a strong default when you want modern instruction following and reasoning without the cost of large models. Ideal for assistants, evaluations, and workflows where efficiency matters.

messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"},
    {"role": "assistant",
     "content": "Sure! Here are some ways to eat bananas and dragonfruits together"},
    {"role": "user", "content": "What about solving an 2x + 3 = 7 equation?"},
]

response = await norman.invoke(
    {
        "model_name": "phi-4-mini-instruct",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": messages
            }
        ]
    }
)