MachinePulseAI/karpo-decision-12B

Karpo Decision 12B

A decision model of the Karpo Decision family: given a state (a web page, a document, a conversation: any text or JSON) and typed questions about it, it returns a probability for every option of every question in one forward pass, without generating text. Fine-tuned from google/gemma-4-12B-it, with a linear readout over the options.

• Parameters: 12.0B in bf16, plus a 255 × 3,840 readout in fp32
• Questions: choice (up to 255 options), noul (true or false), score (a level on a scale)
• Context: 32,768 tokens per sequence; a longer request is split by questions, each part with the full state
• Interfaces: a Python API and an HTTP server in the System One format (POST /v1/systemone)
• Licence: MachinePulse Community License: commercial use is allowed; commercial products and services offered to third parties must show "Powered by karpo-decision (MachinePulse)"; production use for commercial purposes by an entity whose annual revenue, with its affiliates, exceeds US$10M needs a commercial agreement

The Karpo Decision family

Score vs. size on the Decision Index 0.3 public suite, board of 12 November 2026.

Model	Base	Parameters
karpo-decision-31B	Gemma 4 31B	31.3B
karpo-decision-12B	Gemma 4 12B	12.0B
karpo-decision-9B	Qwen3.5 9B	8.4B
karpo-decision-4B	Qwen3.5 4B	4.5B
karpo-decision-2B	Qwen3.5 2B	2.2B
karpo-decision-0.8B	Qwen3.5 0.8B	0.85B

The family's largest model, karpo-decision-31B, is ranked #1 on the Decision Index, with the highest Full score, 73.9, on the maintainers' private tests (edition 0.4.1, board of 12 November 2026, 300 entries); its public index is 75.0.

How it works

• A readout, not generation. Every option is labelled with an option code. At the Answer: ( slot that ends a question's branch, a linear readout scores 255 option codes and a softmax over the question's own options gives the probabilities. Nothing is decoded, so the answer is always one of the options.
• Many questions, one forward pass. A request is one state and any number of questions, encoded as one sequence: the list of questions, the state, then one branch per question with its options.
• Branch attention. A branch reads the shared block (questions and state) and itself, never another branch. The state is computed once per sequence, and no question sees another question's options.

The question-list-first layout follows FlexDM and jev-brain; the shared-block and branch rule follows Sieve-31B-v3.

Use

Inference code: github.com/machinepulse-ai/karpo-decision. Python 3.12 or later and one GPU; the weights take 23.9 GB, long requests more. Loading the checkpoint as a plain chat model does not reproduce its answers.

pip install "karpo-decision @ git+https://github.com/machinepulse-ai/karpo-decision"

from huggingface_hub import snapshot_download
from karpo_decision.decider import Decider

path = snapshot_download("MachinePulseAI/karpo-decision-12B")
decider = Decider.from_pretrained(path, device="cuda")
response = decider.decide({
    "state": "Order 1182: the customer asks to return a blender bought 45 days ago. Policy: returns within 30 days.",
    "questions": {
        "eligible": {"type": "noul", "instructions": "Is the return eligible under the policy?"},
        "action": {"type": "choice", "instructions": "What should the agent do?",
                   "criteria": {"refund": "Issue a full refund", "decline": "Politely decline the return",
                                "escalate": "Escalate to a supervisor"}},
    },
})
print(response["answers"])
# {"eligible": {"type": "noul", "noul": P(true)},
#  "action": {"type": "choice", "choice": <key>, "probabilities": {<key>: p, ...}, "confidence": c}}

HTTP server (POST /v1/systemone, GET /healthz):

CKPT=$(python -c 'from huggingface_hub import snapshot_download; print(snapshot_download("MachinePulseAI/karpo-decision-12B"))')
torchrun --standalone --nproc_per_node 1 -m karpo_decision.serve.server --checkpoint "$CKPT" --port 18440

Decision Index kit, installed separately:

pip install "decision-index @ git+https://github.com/apolinario/decision-index"
python -m decision_index run --engine karpo_decision.serve.engine:DecisionIndexEngine \
    --option checkpoint="$CKPT" --out runs/karpo-decision-12B

Lineage

google/gemma-4-12B → google/gemma-4-12B-it → karpo-decision-12B (fine-tuned)

Limitations

• Text only: the vision and audio embedders are in the checkpoint but unused.
• Not a chat or text-generation model; generation was not evaluated.
• A question that does not fit in 32,768 tokens with its state is refused, never truncated.
• It always picks one of the given options.

Licence

The weights and model files are licensed under the MachinePulse Community License; this summary is for convenience and the LICENSE governs. Commercial use is allowed. Commercial products and services offered to third parties, paid or free, must carry the attribution below; internal use and internal tools, evaluation, testing, benchmarking, research, and pilots of up to 90 days that serve no paying customers need none. Production use for commercial purposes by an entity whose annual revenue, with its affiliates, exceeds US$10M needs a separate commercial agreement: write to contact@machinepulse.ai with the subject line "Karpo Decision licensing: commercial agreement request". Models fine-tuned, merged, quantized, pruned or distilled from these weights stay under the same licence; a model trained only on this model's outputs does not.

karpo-decision-12B is a modified version of google/gemma-4-12B-it (Apache-2.0): decoder weights fine-tuned, readout added. Gemma 4 stays under its own licence, a copy of which is in licenses/Gemma-4-LICENSE; see NOTICE.

The inference code (github.com/machinepulse-ai/karpo-decision) is licensed under Apache-2.0.

Attribution

Commercial products and services offered to third parties that use this model or a derivative must show, on a related website, user interface, about page or product documentation:

Powered by karpo-decision (MachinePulse)

with a link to https://huggingface.co/MachinePulseAI/karpo-decision-12B where the medium allows one.