batiai/ minimax-m2.7:q4

570 Downloads Updated 3 weeks ago

ollama run batiai/minimax-m2.7:q4

curl http://localhost:11434/api/chat \
  -d '{
    "model": "batiai/minimax-m2.7:q4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='batiai/minimax-m2.7:q4',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'batiai/minimax-m2.7:q4',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 3 weeks ago

3 weeks ago

2b9ab1f4f0d7 · 122GB ·

model

archminimax-m2

·

parameters229B

·

quantizationIQ4_XS

122GB

system

You are a helpful AI assistant.

31B

params

{ "num_ctx": 131072, "temperature": 0.7 }

37B

Readme

MiniMax M2.7 — Quantized by BatiAI

Quantized from official MiniMax weights. Verified on real Mac hardware.

Models

Tag	Size	VRAM	M4 Max (128GB)	Use Case
iq3	82GB	104GB	36.7 t/s	128GB Mac

Quick Start

ollama run batiai/minimax-m2.7:iq3

MiniMax M2.7 — Why It Matters

229B Dense — one of the largest open models available
Outperforms GPT-5.3 on GDPval-AA (ELO 1495, highest among open-source)
Toolathon: 46.3% accuracy (global top tier)
Agent Teams, complex Skills, dynamic tool search
Word, Excel, PPT editing with multi-round refinement
Modified-MIT license

Benchmarks — M4 Max (128GB)

Metric	IQ3_XXS
Token gen (short)	22.1 t/s
Token gen (long)	36.7 t/s
Prompt eval	14.8 t/s
VRAM	104 GB (97% GPU)
Cold start	42 seconds
Korean	✅
Tool call JSON	✅

RAM Requirements — Be Honest

Your Mac RAM	IQ3_XXS (82GB)
16GB	❌
32GB	❌
48GB	❌
64GB	❌
96GB	⚠️ Heavy swap
128GB	✅ 36.7 t/s (104GB VRAM)
192GB+	✅ Fast, with headroom

This model requires 128GB+ unified memory. No workarounds — 229B Dense needs real RAM.

For Smaller Macs

Your Mac	Recommended Model
16GB	`batiai/gemma4-e4b:q4` (57 t/s)
24GB	`batiai/gemma4-26b:iq4` (85 t/s)
48GB	`batiai/gemma4-26b:iq4` or `batiai/qwen3.5-35b:iq4`
128GB	`batiai/minimax-m2.7:iq3` (this model)

Why BatiAI?

Quantized directly from official MiniMax weights
IQ3_XXS with imatrix — maximum compression for 128GB Mac
Verified on MacBook Pro M4 Max (128GB)
Korean, tool calling, JSON generation all tested
229B on a laptop — no cloud, no API costs

Built for BatiFlow

Free, on-device AI automation for Mac. 5MB app, 100% local, unlimited.

https://flow.bati.ai

# MiniMax M2.7 — Quantized by BatiAI

Quantized from official MiniMax weights. Verified on real Mac hardware.

## Models

| Tag | Size | VRAM | M4 Max (128GB) | Use Case |
|-----|------|------|----------------|----------|
| **iq3** | **82GB** | **104GB** | **36.7 t/s** | **128GB Mac** |

## Quick Start

```
ollama run batiai/minimax-m2.7:iq3
```

## MiniMax M2.7 — Why It Matters

- **229B Dense** — one of the largest open models available
- Outperforms GPT-5.3 on GDPval-AA (ELO 1495, highest among open-source)
- Toolathon: 46.3% accuracy (global top tier)
- Agent Teams, complex Skills, dynamic tool search
- Word, Excel, PPT editing with multi-round refinement
- Modified-MIT license

## Benchmarks — M4 Max (128GB)

| Metric | IQ3_XXS |
|--------|---------|
| **Token gen (short)** | **22.1 t/s** |
| **Token gen (long)** | **36.7 t/s** |
| Prompt eval | 14.8 t/s |
| VRAM | 104 GB (97% GPU) |
| Cold start | 42 seconds |
| Korean | ✅ |
| Tool call JSON | ✅ |

## RAM Requirements — Be Honest

| Your Mac RAM | IQ3_XXS (82GB) |
|-------------|---------------|
| 16GB | ❌ |
| 32GB | ❌ |
| 48GB | ❌ |
| 64GB | ❌ |
| 96GB | ⚠️ Heavy swap |
| **128GB** | **✅ 36.7 t/s (104GB VRAM)** |
| 192GB+ | ✅ Fast, with headroom |

This model requires **128GB+ unified memory**. No workarounds — 229B Dense needs real RAM.

## For Smaller Macs

| Your Mac | Recommended Model |
|----------|------------------|
| 16GB | `batiai/gemma4-e4b:q4` (57 t/s) |
| 24GB | `batiai/gemma4-26b:iq4` (85 t/s) |
| 48GB | `batiai/gemma4-26b:iq4` or `batiai/qwen3.5-35b:iq4` |
| 128GB | `batiai/minimax-m2.7:iq3` (this model) |

## Why BatiAI?

- Quantized directly from official MiniMax weights
- IQ3_XXS with imatrix — maximum compression for 128GB Mac
- Verified on MacBook Pro M4 Max (128GB)
- Korean, tool calling, JSON generation all tested
- 229B on a laptop — no cloud, no API costs

## Built for BatiFlow

Free, on-device AI automation for Mac. 5MB app, 100% local, unlimited.

https://flow.bati.ai

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