EXAONE-4.0-1.2B

EXAONE-4.0-1.2B (LG AI)

Small, efficient language model focused on instruction following and practical tasks.

• Compact by design. A 1.2B-parameter model from the EXAONE-4.0 family, trained on curated multilingual and instruction data to maximize usefulness at small scale.

• Task-oriented behavior. Performs well on chat, summarization, classification, and simple reasoning compared to other models in the ~1B range.

• Low-cost inference. Small memory footprint and fast generation make it suitable for CPUs, edge deployments, and high-throughput services.

• Straightforward setup. Single model, standard tokenizer, no special tooling or pipelines required.

• Open and usable. Released under a permissive open license, safe for commercial and research use.

Why pick it for Norman AI?

EXAONE-4.0-1.2B is a good fit when you need an ultra-light LLM that’s still instruction-aware. Ideal for routing, lightweight assistants, and experiments where latency and cost matter most.

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": "exaone-4.0-1.2b",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": messages
            }
        ]
    }
)