rubinmaximilian/ Monk-Router-Gemma4e2b

13 Downloads Updated 2 hours ago

A low-latency agentic router powered by the brand-new Gemma 4 E2B (Effective 2B) architecture. Optimized for the Monk AI framework to provide near-instant task delegation and "Edge-to-Business" tool-calling on Jetson hardware.

vision tools thinking audio

ollama run rubinmaximilian/Monk-Router-Gemma4e2b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "rubinmaximilian/Monk-Router-Gemma4e2b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='rubinmaximilian/Monk-Router-Gemma4e2b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'rubinmaximilian/Monk-Router-Gemma4e2b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code

Claude Code ollama launch claude --model rubinmaximilian/Monk-Router-Gemma4e2b

Codex

Codex ollama launch codex --model rubinmaximilian/Monk-Router-Gemma4e2b

OpenCode

OpenCode ollama launch opencode --model rubinmaximilian/Monk-Router-Gemma4e2b

OpenClaw

OpenClaw ollama launch openclaw --model rubinmaximilian/Monk-Router-Gemma4e2b

Models

Name

1 model

Size

Context

Input

Monk-Router-Gemma4e2b:latest

7.2GB · 128K context window · Text, Image · 2 hours ago

Monk-Router-Gemma4e2b:latest

7.2GB

128K

Text, Image

Readme

Monk-Router-gemma4e2b

Monk-Router-gemma4e2b is a performance-first router designed for the Monk AI assistant. Leveraging Google’s 2026 E2B (Effective 2 Billion) architecture, it offers the fastest possible decision-making speed for edge computing environments (versus the similar model based on Phi4-mini).

This model is ideal for users prioritizing latency and VRAM efficiency on devices like the Jetson Orin Nano or MacBook Air. Please let me know if I should make an even larger model for scaled applications!

Performance

Low VRAM Footprint: Uses ~1.5GB of VRAM, allowing it to stay resident while worker models load.
Agentic Efficiency: Built on Gemma 4’s native tool-calling distillation for superior instruction following.
Speed: Optimized for “Time to First Token” (TTFT) in real-time assistant workflows.

Logic Thresholds

Edge Route: Simple queries, small code snippets (<100 lines), and general chat.
GPU Route: High-VRAM requirements, multi-file analysis, and thermal-intensive tasks.

Example Output

”`json { “logic”: “General logic task. Keeping on local Jetson.”, “tool_call”: { “name”: “switch_model”, “parameters”: { “model_name”: “gemma4-e2b” } } }

# Monk-Router-gemma4e2b

**Monk-Router-gemma4e2b** is a performance-first router designed for the **Monk AI** assistant. Leveraging Google's 2026 **E2B (Effective 2 Billion)** architecture, it offers the fastest possible decision-making speed for edge computing environments ([versus the similar model based on Phi4-mini](https://ollama.com/rubinmaximilian/Monk-Router-phi4mini)).

This model is ideal for users prioritizing **latency** and **VRAM efficiency** on devices like the Jetson Orin Nano or MacBook Air. Please let me know if I should make an even larger model for scaled applications!

### Performance
- **Low VRAM Footprint:** Uses ~1.5GB of VRAM, allowing it to stay resident while worker models load.
- **Agentic Efficiency:** Built on Gemma 4's native tool-calling distillation for superior instruction following.
- **Speed:** Optimized for "Time to First Token" (TTFT) in real-time assistant workflows.

### Logic Thresholds
- **Edge Route:** Simple queries, small code snippets (<100 lines), and general chat.
- **GPU Route:** High-VRAM requirements, multi-file analysis, and thermal-intensive tasks.

### Example Output
```json
{
  "logic": "General logic task. Keeping on local Jetson.",
  "tool_call": {
    "name": "switch_model",
    "parameters": { "model_name": "gemma4-e2b" }
  }
}

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