Kev 4B

Kev 4B (jaredpalmer/kev-4b) is an open-weight decision model from Jared Palmer: a LoRA adapter and pointer head on Qwen3.5-4B-Base, served over the same /v1/systemone contract as Jev. It generates no text — it returns calibrated probabilities for typed questions in a single forward pass. The recommended checkpoint of the Kev family (0.8B, 4B, 9B), Apache-2.0 licensed. 8,192-token context.

When to pick Kev over Jev

• Cheapest possible gating — same pricing (input only, $0.042/1M) on a lighter model: for high-volume checks (“is this tool call safe?”, “which team owns this ticket?”) quality is usually on par.
• Open-source requirements — Apache-2.0: weights, code and eval suites are public (github.com/jaredpalmer/kev).
• Short states — 8K context vs Jev’s 32K: if your state fits, Kev is enough.

Pick Jev when the state is large or you need maximum distribution calibration.

Quick start

curl

curl https://api.mixen.ai/v1/decisions \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kev",
    "state": "Task: clean up inactive accounts before the quarterly report. Proposed tool call: delete_rows(table=\"customers\", where=\"last_login < 2023-01-01\"). Context: the customers table has 48,210 rows and no backup was taken today.",
    "questions": {
      "safe_to_run": {
        "type": "noul",
        "instructions": "Is this action safe to run without a human approving it first?",
        "criteria": {
          "true": "Reversible or low-impact, and clearly within the stated task.",
          "false": "Destructive, irreversible, or broader than the task requires."
        }
      }
    }
  }'

Python

import requests

r = requests.post(
    "https://api.mixen.ai/v1/decisions",
    headers={"Authorization": f"Bearer {KEY}"},
    json={
        "model": "kev",
        "state": "Task: clean up inactive accounts. Tool call: delete_rows(...)",
        "questions": {
            "safe_to_run": {
                "type": "noul",
                "instructions": "Is this action safe without human approval?",
                "criteria": {
                    "true": "Reversible, low-impact, within the task.",
                    "false": "Destructive, irreversible, or broader than the task.",
                },
            }
        },
    },
)
if r.json()["answers"]["safe_to_run"]["noul"] < 0.8:
    escalate_to_human()

Pricing

Same as Jev: only input tokens are billed (state plus questions); output is free. A typical request is a few hundred tokens — a fraction of a cent. Charging follows the actual usage.cost from the response.

Limitations

• 8K-token context.
• Same primitives as Jev: noul, choice, score — full schema in the Decisions guide.