Skip to main content
Aider is a terminal-based AI coding assistant that works directly with your git repository. It can make changes across multiple files, understand your codebase context, and commit changes automatically.
Verified working with MiniMax-M2.5 on Infercom.

​
Prerequisites

​
Installation

pip3 install aider-chat
Verify installation: 
aider --version

​
Configuration

Set these environment variables: 
export OPENAI_API_BASE="https://api.infercom.ai/v1"
export OPENAI_API_KEY="your-infercom-api-key"
Then run Aider: 
aider --model openai/MiniMax-M2.5
Add the export commands to your shell profile (~/.bashrc, ~/.zshrc) for persistence.

​
Option 2: Config File

Create ~/.aider.conf.yml: 
openai-api-base: https://api.infercom.ai/v1
openai-api-key: your-infercom-api-key
model: openai/MiniMax-M2.5
Then simply run: 
aider

​
Option 3: Command Line

Pass everything on the command line: 
OPENAI_API_BASE="https://api.infercom.ai/v1" \
OPENAI_API_KEY="your-infercom-api-key" \
aider --model openai/MiniMax-M2.5

​
Model

Use MiniMax-M2.5 with the openai/ prefix: 
openai/MiniMax-M2.5
The openai/ prefix is required. Using just MiniMax-M2.5 will not work.

​
Example Usage

​
Interactive Mode

cd your-project
aider --model openai/MiniMax-M2.5
Then chat with Aider: 
> Add error handling to the main function in app.py

​
Non-Interactive Mode

aider --model openai/MiniMax-M2.5 \
  --message "Add a hello world function" \
  --yes \
  main.py

​
Disable Auto-Commits

aider --model openai/MiniMax-M2.5 --no-auto-commits

​
Reasoning Output

MiniMax-M2.5 has built-in reasoning capabilities. To see the model’s thinking process, disable streaming: 
aider --model openai/MiniMax-M2.5 --no-stream
You’ll see a THINKING section before each response showing the model’s reasoning.

​
Controlling Reasoning Effort

You can adjust reasoning intensity with --reasoning-effort: 
aider --model openai/MiniMax-M2.5 \
  --no-stream \
  --reasoning-effort medium \
  --no-check-model-accepts-settings
Valid values: low, medium, high
--no-stream is required to display reasoning output. The --no-check-model-accepts-settings flag is needed because Aider doesn’t recognize MiniMax’s reasoning support by default.

​
Troubleshooting

​
Unknown Model Warning

You may see this warning: 
Warning for openai/MiniMax-M2.5: Unknown context window size and costs,
using sane defaults.
Fix: Create ~/.aider.model.metadata.json with MiniMax-M2.5 specs: 
{
  "openai/MiniMax-M2.5": {
    "max_tokens": 16384,
    "max_input_tokens": 160000,
    "max_output_tokens": 16384,
    "input_cost_per_token": 0.0000003,
    "output_cost_per_token": 0.0000012,
    "litellm_provider": "openai",
    "mode": "chat",
    "accepts_settings": ["reasoning_effort"]
  }
}
This enables proper context window handling and cost tracking. Alternative: Suppress the warning with --no-show-model-warnings: 
aider --model openai/MiniMax-M2.5 --no-show-model-warnings

​
Connection Errors

Verify your configuration: 
curl -s https://api.infercom.ai/v1/models \
  -H "Authorization: Bearer $OPENAI_API_KEY" | head -20
You should see a list of available models.

​
Model Not Found

Ensure you’re using the correct model name with the openai/ prefix: 
# Correct
aider --model openai/MiniMax-M2.5

# Wrong
aider --model MiniMax-M2.5

​
Performance

  • Token throughput: 400+ tokens/sec with MiniMax-M2.5
  • Context window: 160K tokens (163,840) - handles large codebases
  • Quality: 75.8% SWE-bench Verified

​
Next Steps