Havenlon/ Execution-Boundary-Qwen35-4B-Q4_K_M

93 Downloads Updated 1 week ago

A Havenlon-focused Qwen3.5 4B model balancing local deployment cost with execution-boundary reasoning, Adversarial Completeness, and AI Agent execution control.

ollama run Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M

curl http://localhost:11434/api/chat \
  -d '{
    "model": "Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

Execution-Boundary-Qwen35-4B-Q4_K_M:latest

2.7GB · 256K context window · Text · 1 week ago

Execution-Boundary-Qwen35-4B-Q4_K_M:latest

2.7GB

256K

Text

Readme

Execution-Boundary-Qwen35-4B-Q4_K_M

A Havenlon-focused Qwen3.5 4B model for execution-boundary reasoning, Adversarial Completeness, execution control, and AI Agent execution risk.

The 4B model is designed as a balance between compact deployment and stronger domain reasoning.

Focus

Execution Control
Execution Boundaries
Adversarial Completeness
Intent-to-Execution consistency
Policy vs. Final Authority
Independent veto capability
Evidence chains
Hardware-backed execution boundaries
AI Agent execution risk
Fail-Secure behavior

Run

ollama run Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M

Quantization

Q4_K_M

Core Perspective

Havenlon does not treat approval, authentication, policy, or software intent as equivalent to final execution authority.

A request may be valid and still be unsafe to execute.

A secure execution boundary must therefore be able to independently reject an action when critical facts are missing, inconsistent, expired, conflicting, or unverified.

Intent is not Execution.

Approval is not Execution Control.

Policy is not Final Authority.

A system must preserve the ability to refuse execution.

# Execution-Boundary-Qwen35-4B-Q4_K_M

A Havenlon-focused Qwen3.5 4B model for execution-boundary reasoning, Adversarial Completeness, execution control, and AI Agent execution risk.

The 4B model is designed as a balance between compact deployment and stronger domain reasoning.

## Focus

- Execution Control
- Execution Boundaries
- Adversarial Completeness
- Intent-to-Execution consistency
- Policy vs. Final Authority
- Independent veto capability
- Evidence chains
- Hardware-backed execution boundaries
- AI Agent execution risk
- Fail-Secure behavior

## Run

```bash
ollama run Havenlon/Execution-Boundary-Qwen35-4B-Q4_K_M
```

## Quantization

Q4_K_M

## Core Perspective

Havenlon does not treat approval, authentication, policy, or software intent as equivalent to final execution authority.

A request may be valid and still be unsafe to execute.

A secure execution boundary must therefore be able to independently reject an action when critical facts are missing, inconsistent, expired, conflicting, or unverified.

> Intent is not Execution.

> Approval is not Execution Control.

> Policy is not Final Authority.

> A system must preserve the ability to refuse execution.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)