Happy Horse 1.0 is now on ModelsLab

Try Now
Skip to main content
Available now on ModelsLab · Language Model

MiniMax: MiniMax M2.7Self-Evolves. Outperforms Giants

Try MiniMax: MiniMax M2.7API Documentation

Build Agents That Improve

Self-Evolving Core

Autonomously Optimizes Itself

Runs 100+ scaffold iterations, handles 30-50% of RL workflows without humans.

Agentic Power

Masters Complex Harnesses

Builds agent teams, dynamic tools, 97% skill compliance on 40+ complex tasks.

Efficient Inference

10B Params, Tier-1 Scores

Activates 10B parameters for 56% SWE-Pro, 100 TPS, $0.30/M input via MiniMax: MiniMax M2.7 API.

Examples

See what MiniMax: MiniMax M2.7 can create

Copy any prompt below and try it yourself in the playground.

Code Scaffold

“Design an agent harness in Python using OpenClaw framework to optimize reinforcement learning experiments. Include memory updates, skill building for 40 complex tasks over 2000 tokens each, and self-evaluation loops for 100 iterations.”

Workflow Debug

“Analyze this failing ML pipeline code, identify root causes, propose fixes, and generate an improved version with agentic multi-step reasoning for production deployment.”

Financial Model

“Build Excel-compatible financial model for revenue forecasting using historical data. Include sensitivity analysis, Monte Carlo simulations, and export to spreadsheet format.”

Document Pipeline

“Generate full technical report on AI agent benchmarks in Word format. Cover SWE-Pro 56%, Terminal Bench, include charts, executive summary, and self-optimization recommendations.”

For Developers

A few lines of code.
Agents Evolve. Code Ships.

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

  • Serverless: scales to zero, scales to millions
  • Pay per token, no minimums
  • Python and JavaScript SDKs, plus REST API
API Documentation 
import requests


response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
  "key": "YOUR_API_KEY",
  "prompt": "",
  "model_id": ""
}
)
print(response.json())

FAQ

Common questions about MiniMax: MiniMax M2.7

Read the docs

MiniMax: MiniMax M2.7 is a 10B parameter LLM with self-evolution, running 100+ optimization cycles. It rivals larger models on SWE-Pro at 56%. Access via MiniMax: MiniMax M2.7 API.

It updates memory, builds skills, and refines harnesses autonomously. Handles 30-50% RL workflows and won 9 ML competition golds. Ties Gemini on MLE-Bench.

Costs $0.30 per million input tokens. Delivers 46-100 TPS with 200k context. Suitable for agentic tasks via minimax: minimax m2.7 endpoints.

Scores 56% on SWE-Pro, 57% Terminal Bench. Excels in AI coding tools, multi-turn dialogue. Use MiniMax: MiniMax M2.7 model for production pipelines.

Matches Opus-level benchmarks at 50x lower cost, 3x speed. Self-evolving agents make it ideal alternative. Try MiniMax: MiniMax M2.7 LLM for efficiency.

Supports 205k token context window for long conversations. Handles complex agent harnesses and document generation. Integrates via standard LLM endpoints.

Ready to create?

Start generating with MiniMax: MiniMax M2.7 on ModelsLab.

Try MiniMax: MiniMax M2.7API Documentation