Lumma-Fev-0.6B

Lumma-Fev-0.6B is a decision model. It takes one document (the state) and a set of typed questions, and returns a probability distribution for each question in one forward pass. It does not generate text, so there is nothing to parse and nothing to hallucinate.

It is a Jev-like model, fine-tuned on top of our Lumma-0.6B-Base, which we pre-trained from scratch.

🚀 Coming this week: Lumma-Fev-4B and Lumma-Fev-9B, released with a detailed blog covering full benchmarks.

Model details

Developed by	FrontiersMind
Model type	Causal transformer, prefill-only, block-causal branch mask, pointer readout
Base model	Lumma-0.6B-Base (649M parameters, frozen)
Precision	bf16 (training and serving on H200)
Question types	noul (yes/no), choice (2–255 options), score (2–255 ordered levels)
Language	English

Known limits

• English only.
• Fit a temperature on your own data before gating automation on the probabilities.

🚀 Usage

import torch
import json
from transformers import AutoModel, AutoTokenizer

model_id = "FrontiersMind/Lumma-fev-0.6b"
tok = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModel.from_pretrained(model_id, trust_remote_code=True, dtype=torch.bfloat16)

response = model.decide(
    state="I was charged twice. Please refund one.",
    questions={
        "billing": {
            "type": "noul",
            "instructions": "Is this about billing?"
        },
        "team": {
            "type": "choice",
            "instructions": "Which team should handle this?",
            "criteria": {
                "billing": "Payments and refunds",
                "shipping": "Delivery problems"
            },
        },
    },
)

print(json.dumps(response, indent=4))

📬 Feedback & Suggestions

We'd love to hear your thoughts, feedback, and ideas!

• Discord: https://discord.gg/ZGdjCdRt
• Email: support@frontiersmind.ai
• Official Website: https://www.frontiersmind.ai/
• LinkedIn: https://www.linkedin.com/company/frontiersmind/
• X (Twitter): https://x.com/FrontiersMind