Happy Horse 1.0 is now on ModelsLab

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

Minimax M1 80KReason Deep. Context Vast

Try Minimax M1 80KAPI Documentation

Scale Reasoning Efficiently

Hybrid MoE

456B Parameters Active 46B

Activates 45.9B parameters per token in Mixture-of-Experts for efficient complex reasoning.

Lightning Attention

1M Token Context

Processes 1 million input tokens with 80K output using hybrid attention for long documents.

Test-Time Scaling

80K Thinking Budget

Extends compute for superior SWE-bench (56%) and AIME 2024 (86%) performance.

Examples

See what Minimax M1 80K can create

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

Code Debug

“Analyze this 50K token Python codebase with bugs in the async handler. Step through logic, identify issues in dependency injection and error handling, then output fixed code with explanations.”

Document Summary

“Summarize key insights from this 800K token technical report on quantum computing advancements, highlighting breakthroughs in error correction and scalability challenges.”

Math Proof

“Prove the Riemann hypothesis implications for prime distribution using chain-of-thought reasoning over extended steps, citing relevant theorems and counterexamples.”

Agent Workflow

“Design multi-step agent plan to optimize supply chain logistics from this 200K token dataset, incorporating tool calls for inventory query and route optimization.”

For Developers

A few lines of code.
Reasoning chains. One call.

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 M1 80K

Read the docs

Minimax M1 80K is a 456B parameter hybrid MoE reasoning model with 80K thinking budget. It supports 1M token context via lightning attention. Optimized for software engineering and long-context tasks.

Send requests to Minimax M1 80K API in system/user/assistant format for chain-of-thought reasoning. Handles up to 1M input and 80K output tokens. Uses vLLM or SGLang for deployment.

Supports 1 million token context window for extensive document processing. Enables 80K reasoning output with 30% compute of DeepSeek R1. Outperforms on long-context benchmarks.

Achieves 56% on SWE-bench Verified, outperforming most open models. Excels in software engineering and tool use. Suitable for advanced coding with extensive reasoning.

Minimax M1 80K uses less FLOPs at 100K generation and has 8x larger context. Scores close on SWE-bench (56% vs 57.6%). Leads open models in agentic tasks.

Consider MiniMax-M1-40k for lighter loads or Qwen3-235B. Minimax M1 80K excels in reasoning depth. Evaluate via benchmarks like AIME or TAU-bench for fit.

Ready to create?

Start generating with Minimax M1 80K on ModelsLab.

Try Minimax M1 80KAPI Documentation