Instructions to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("baa-ai/MiniMax-M2.7-RAM-120GB-MLX")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default baa-ai/MiniMax-M2.7-RAM-120GB-MLX

Run Hermes

hermes

OpenClaw new

How to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "baa-ai/MiniMax-M2.7-RAM-120GB-MLX" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

MLX LM

How to use baa-ai/MiniMax-M2.7-RAM-120GB-MLX with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "baa-ai/MiniMax-M2.7-RAM-120GB-MLX"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "baa-ai/MiniMax-M2.7-RAM-120GB-MLX",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

MiniMax-M2.7 on mlx_lm.server loads fine and the first API/tool call succeeds

by jarvisfoobi - opened Apr 15

Discussion

jarvisfoobi

Apr 15

The model loads fine and the first API/tool-like call succeeds.
But in a multi-step workflow, the model corrupts identifiers between steps.

Example:

first response contains correct issue ID:
6e5bb523-70d4-4fac-a81a-eb0e2864a20e
next generated command uses:
6e5bb523-70d4-4fac-a81a-eb0e28a64a20e

So one character group changes, and the next API call fails with:
{"error":"Issue not found"}

This does not look like:

model load failure
missing template/tokenizer/config
wrong base URL
failed first call

It looks like:

argument / ID corruption across sequential calls

Question:
Has anyone seen MiniMax-M2.7 on MLX corrupt IDs / tool-call arguments like this in multi-step agent workflows?

If needed, I can share:

the exact first successful call
the next corrupted call
the MLX server settings and chat template setup

tomkay

baa.ai org Apr 15

Yeah typical tokenization drift.

The only way around it is to reduce temperature (temp=0) and top_p=1 for those specific tool calls.

There is also a known issue with mlx-lm and MiniMax (see MLX-LM Issue #1145) where the model sometimes omits angle brackets in tool tags. This "messy" output can confuse the model's internal state for the next turn.

To address that you want to ensure your chat_template or system prompt explicitly reinforces the XML structure

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment