MiniMax: MiniMax-M2.7 API
MiniMax-M2.7 is a frontier-class model specifically engineered for autonomous recursive self-evolution and complex multi-agent orchestration. Building on the ultra-efficient MoE architecture of the M2 series (230B total parameters, 10B active), M2.7 introduces an "Internal Harness" system that allows the model to autonomously evaluate its failure paths, modify its own scaffold code, and iterate on reinforcement learning (RL) processes. It excels in professional software engineering, scoring 56.22% on the SWE-Pro benchmark, matching the capabilities of GPT-5.3-Codex. This model is designed to handle high-fidelity office document editing (Word, Excel, PPT) and maintains a 97% adherence rate even when invoking over 40 complex skills in a single session.
- Context window: 204,800 tokens
- Input: text, image
- Output: text
- Reasoning: Supported
- Tool calling: Supported
- File input: Supported
- Released: 2026-03-17
- Knowledge cutoff: 2026-02-28
Frequently Asked Questions
What is the context window of MiniMax-M2.7?
MiniMax-M2.7 supports a context window of up to 204,800 tokens.
Does MiniMax-M2.7 support function calling?
Yes. MiniMax-M2.7 supports tool / function calling.
Does MiniMax-M2.7 support reasoning?
Yes. MiniMax-M2.7 is a reasoning-capable model.
What is the knowledge cutoff of MiniMax-M2.7?
The knowledge cutoff of MiniMax-M2.7 is 2026-02-28.