Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
AIHubMix logo

Model details

MiniMax M2.7

MiniMax M2.7 sits inside MiniMax's M-series text lineup, presented on the provider's official model page between the flagship M3 and the earlier M2.5. Its product framing emphasizes "model self-improvement" and productivity innovation, signaling an intent to push iterative capability gains rather than market a single breakthrough. The model's substantial 229B-parameter scale, as documented by its Ollama library entry, points to a large general-purpose architecture aimed at handling complex reasoning chains, code synthesis, and long-running tool-driven tasks.

In practice, M2.7 is positioned for coding and agentic workflows where the model can plan, call tools, and sustain extended sessions. Ollama describes it as MiniMax's M2-series model "for coding, agentic workflows, and professional productivity," and the listing shows a 200K-token context window that supports lengthy codebases, multi-document reasoning, and long agent traces. Integration entry points such as Claude Code, OpenCode, and Hermes Agent via the Ollama launch command illustrate how developers can drop the model into existing agent harnesses, while 2.4M reported downloads reflect early traction through that distribution channel.

AIHubMixminimax-m2.7minimax

Quick Info

Powered by
Provider
AIHubMix
Model key
minimax-m2.7
Release date
Mar 18, 2026
Last updated
Mar 18, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.30
Output token cost
$1.20

Limits

Output tokens
128,000 tokens
Context window
204,800 tokens

Transparent token rates

Compare MiniMax M2.7 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about MiniMax M2.7

Vercel AI Gateway

CoverageBenchmark

An InferenceX architecture page documents the M2.7 model specification and its position within the M2 series. The architecture is recorded as a 230B-parameter MoE with 9.8B activated per token, a 197K context window, 62 layers of grouped query attention with Top-8/256 experts, RMSNorm, RoPE, multi-token prediction acro Positioning is framed as agentic and oriented toward coding and office-work tasks, with M2.7 described as the series' first model deeply participating in its own evolution, able to build complex agent harnesses and complete elaborate productivity tasks using Agent Teams, complex Skills, and dynamic tool search. The mod

Videos about MiniMax M2.7

More models around MiniMax M2.7