MiniMax-M2.1

MiniMax-M2.1 sits within the MiniMax family of large language models as a release dedicated to coding workflows, with the official news announcement positioning it around significantly enhanced multi-language programming and complex real-world tasks. The companion Models documentation page lists a MiniMax-M2.1 entry describing the model as having 230B total parameters with 10B activated per inference and being optimized for code generation and refactoring, highlighting polyglot code mastery and precision code refactoring among its intended strengths. Together these signals frame M2.1 as a sparsely activated, code-centric model designed to handle diverse programming languages and intricate engineering scenarios rather than as a general-purpose chat model.

In practical terms, M2.1 is best understood as a specialist coding variant within the MiniMax line: it targets multilingual code generation, refactoring, and the kind of structured engineering work that benefits from a large parameter pool paired with efficient per-query activation. The Models documentation places M2.1 alongside higher-numbered siblings and a separate "highspeed" alias family, suggesting a tier where M2.1 emphasizes depth and correctness in code tasks while speed-optimized variants exist for latency-sensitive deployments. For teams selecting a model in this family, M2.1 is a natural fit when complex, multi-language codebases and careful refactoring matter more than the latest frontier multimodal features found in newer MiniMax releases.

Kilo Gatewayminimax/minimax-m2.1minimax

Open provider page Provider docs

Quick Info

Provider: Kilo Gateway
Model key: minimax/minimax-m2.1
Release date: Dec 23, 2025
Last updated: Dec 23, 2025
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.30
Output token cost: $1.20

Limits

Output tokens: 131,072 tokens
Context window: 204,800 tokens

Transparent token rates

Compare MiniMax-M2.1 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-M2.1

MiniMax Token Plan (minimax.cn)

Coverage

The Rundown AI's independent tool overview (August 28, 2026) profiles MiniMax-M2.1 as a 230-billion-parameter Mixture-of-Experts model with roughly 10 billion active parameters, released by MiniMax on December 23, 2025 with a 204,800-token API context window. The article emphasizes training focused on non-Python langua The overview frames M2.1 as a now-legacy model: MiniMax-M2.7 is the current M2-series text endpoint, and M3 adds a one-million-token context window, native image and video input, and computer use. The standard and high-speed API endpoints for M2.1 remain available but are listed under legacy pricing. The page recommend

Kilo Gateway

CoverageBenchmark

Artificial Analysis provides quantitative analysis of MiniMax-M2.1, confirming it as an open-weights model with 230 billion total parameters and 10 billion active, a 205k token context window, and MIT licensing with weights on Hugging Face. The model scores 21 on the Artificial Analysis Intelligence Index (above averag Pricing data shows M2.1 at $0.30 per million input tokens and $1.20 per million output tokens with a 90% cache discount. The page flags M2.1 as deprecated for ongoing benchmarking, noting that MiniMax has launched MiniMax-M2.5 as a newer successor and recommending users consider it instead. Input and output modalities

Moark

Coverage

MiniMax announced MiniMax M2.1 on December 23, 2025, as an update focused on real-world complex tasks rather than pure cost/accessibility. The official post states that, unlike many models optimized primarily for Python, M2.1 has been systematically enhanced across Rust, Java, Golang, C++, Kotlin, Objective-C, TypeScri The same announcement highlights a leap in WebDev and AppDev capability and aesthetics, noting that M2.1 addresses the widely recognized weakness in mobile development across the industry by significantly strengthening native Android and iOS capabilities. MiniMax also reports systematic enhancement of the model's desig

OpenCode Zen

CoverageRelease Notes

Opper AI's release tracker lists MiniMax M2.1 as released on 23 Dec 2025 with a 205K context window, API pricing of $0.30 per million input tokens and $1.20 per million output tokens, and an Artificial Analysis intelligence score of 32. The tracker places M2.1 in the broader MiniMax M-series timeline between M2 (releas The same tracker shows that 6 of 7 tracked MiniMax model releases are runnable on Opper, situating M2.1 within a continuing release cadence that includes M2.1, M2.5, M2.7, and the M3 flagship. The 23 Dec 2025 date and 205K context figure align with other independent reporting and serve as a cross-checkable reference po