MiniMax
MiniMax M3
MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which replaces full attention with KV-block selection to cut per-token compute at long context — roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill and decode while retaining quality across most tasks. Trained as a native multimodal model on interleaved data and tuned for multi-turn, production-like collaboration via an interactive user-simulator framework, the model is oriented toward sustained, multi-step tasks rather than single-turn execution.
MiniMax M2.7
MiniMax-M2.5 is a SOTA large language model designed for real-world productivity. Trained in a diverse range of complex real-world digital working environments, M2.5 builds upon the coding expertise of M2.1 to extend into general office work, reaching fluency in generating and operating Word, Excel, and Powerpoint files, context switching between diverse software environments, and working across different agent and human teams. Scoring 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench, and 76.3% on BrowseComp, M2.5 is also more token efficient than previous generations, having been trained to optimize its actions and output through planning.
MiniMax M1
MiniMax-M1: The World's First Open-Weight, Large-Scale Hybrid Attention Inference Model MiniMax-M1 adopts a Mixture of Experts (MoE) architecture and integrates the Flash Attention mechanism. The model contains a total of 456 billion parameters, with 45.9 billion parameters activated per token. Natively, the M1 model supports a context length of 1 million tokens—8 times that of DeepSeek R1. Additionally, by combining the CISPO algorithm with an efficient hybrid attention design for reinforcement learning training, MiniMax-M1 achieves industry-leading performance in long-context reasoning and real-world software engineering scenarios.
Minimax M2.1
MiniMax M2.1 is a cutting-edge AI model designed to revolutionize how developers build software. With enhanced multi-language programming support, it excels at generating high-quality code across popular languages such as Rust, Java, Go, C++, Kotlin, Objective-C, TypeScript, and JavaScript. Key improvements include: 22% faster response times and 30% lower token consumption for efficient workflows. Seamless integration with leading development frameworks (Claude Code, Droid Factory AI, BlackBox, etc.). Full-stack development capabilities, from mobile (Android/iOS) to web and 3D interactive prototyping. Optimized performance-to-cost ratio, making AI-assisted development more accessible. Whether you're a software engineer, app developer, or tech innovator, M2.1 empowers smarter coding with industry-leading AI.
MiniMax M2.5
MiniMax M2.1 is a cutting-edge AI model designed to revolutionize how developers build software. With enhanced multi-language programming support, it excels at generating high-quality code across popular languages such as Rust, Java, Go, C++, Kotlin, Objective-C, TypeScript, and JavaScript. Key improvements include: 22% faster response times and 30% lower token consumption for efficient workflows. Seamless integration with leading development frameworks (Claude Code, Droid Factory AI, BlackBox, etc.). Full-stack development capabilities, from mobile (Android/iOS) to web and 3D interactive prototyping. Optimized performance-to-cost ratio, making AI-assisted development more accessible. Whether you're a software engineer, app developer, or tech innovator, M2.1 empowers smarter coding with industry-leading AI.
MiniMax M2.7-highspeed
MiniMax M2.7-highspeed is an accelerated SOTA model engineered for scenarios demanding extreme efficiency. It perfectly inherits the core intelligence and robust digital workspace capabilities of the standard M2.7. In real-world software engineering, M2.7 excels by independently driving end-to-end project delivery while efficiently handling advanced tasks such as log analysis, bug troubleshooting, code security, and machine learning. In the professional workspace, it boasts the highest open-source GDPval-AA score (1495 ELO). It delivers high-fidelity, complex editing and multi-turn revisions across the Office suite (Excel, PPT, Word), elevating task execution to industry-leading standards. Built for complex environment interactions, M2.7 maintains an impressive 97% skill-following rate even with complex, long-context tool calls (>2000 tokens). Beyond its robust productivity, M2.7 breaks the "cold tool" stereotype of traditional models. With exceptional identity retention and high emotional intelligence (EQ),
MiniMax M2.5-highspeed
MiniMax M2.5-highspeed is an accelerated SOTA model engineered for scenarios demanding extreme efficiency. It perfectly inherits the core intelligence and robust digital workspace capabilities of the standard M2.5—including its 80.2% score on SWE-Bench Verified, seamless manipulation of Office documents, and versatility in cross-software collaboration. With zero compromise on reasoning precision or logical depth, the Highspeed version delivers ultra-low latency inference through rigorous engineering optimization. This means you get more than just an intelligent assistant capable of planning and self-optimization; you gain a "high-velocity engine" that responds to high-frequency calls and processes complex document streams in near real-time, making it ideal for latency-sensitive interactive applications and large-scale automated pipelines.
MiniMax H3-Context-IR
MiniMax H3-Context-IR native prompt enhancement (token-metered)
MiniMax H3 Regeneration
MiniMax H3 native video regeneration (768P-spec source re-generated to 2K)
MiniMax H3 Max
MiniMax H3
MiniMax 快速复刻
使用 MiniMax 声音克隆能力从音频 URL 复刻声音。平台会将源音频上传到 MiniMax，创建可复用的 voice_id，返回试听音频 URL，并通过短文本 TTS 请求预热克隆声音。
MiniMax Sound Design
Generate a personalized voice based on a text description. Returns a `voice_id` that can be used with the T2A text-to-speech API, along with a hex-encoded preview audio sample.
MiniMax Speech 2.8 Turbo 异步语音生成
The Minimax series offers reliable synthesis capabilities, making it ideal for production environments. Designed for production-level use, this series prioritizes stability and predictable output. The speech synthesis supports multiple languages and emotional tone control, making it suitable for voiceovers, announcements, customer service, and character dialogue. The real-time inference API delivers stable performance with no waiting time and is affordably priced.
MiniMax Speech 2.8 HD 异步语音生成
The Minimax series offers reliable synthesis capabilities, making it ideal for production environments. Designed for production-level use, this series prioritizes stability and predictable output. The speech synthesis supports multiple languages and emotional tone control, making it suitable for voiceovers, announcements, customer service, and character dialogue. The real-time inference API delivers stable performance with no waiting time and is affordably priced.
MiniMax Speech 2.8 Turbo 同步语音生成
The Minimax series offers reliable synthesis capabilities, making it ideal for production environments. Designed for production-level use, this series prioritizes stability and predictable output. The speech synthesis supports multiple languages and emotional tone control, making it suitable for voiceovers, announcements, customer service, and character dialogue. The real-time inference API delivers stable performance with no waiting time and is affordably priced.
MiniMax Speech 2.8 HD Sync 同步语音生成
The Minimax series offers reliable synthesis capabilities, making it ideal for production environments. Designed for production-level use, this series prioritizes stability and predictable output. The speech synthesis supports multiple languages and emotional tone control, making it suitable for voiceovers, announcements, customer service, and character dialogue. The real-time inference API delivers stable performance with no waiting time and is affordably priced.