MiniMax 推出 MiniMax-01 系列，这是一个新的基础模型系列，旨在处理超长上下文并增强 AI 代理开发

如今，MiniMax 在美国可能最为人所知，因为它是 Hailuo 背后的一家新加坡公司，Hailuo 是一种逼真的高分辨率生成 AI 视频模型，可与 Runway、OpenAI 的 Sora 和 Luma AI 的 Dream Machine 竞争。

MiniMax, a Singaporean company, is perhaps best known in the U.S. for its realistic, high-resolution generative AI video model Hailuo, which competes with Runway, OpenAI’s Sora and Luma AI’s Dream Machine.

MiniMax 是一家新加坡公司，在美国最出名的可能是其逼真的高分辨率生成 AI 视频模型 Hailuo，该模型与 Runway、OpenAI 的 Sora 和 Luma AI 的 Dream Machine 竞争。

But MiniMax has far more tricks up its sleeve: Today, for instance, it announced the release and open-sourcing of the MiniMax-01 series, a new family of models built to handle ultra-long contexts and enhance AI agent development.

但 MiniMax 还有更多的技巧：例如，今天，它宣布发布并开源 MiniMax-01 系列，这是一个新的模型系列，旨在处理超长上下文并增强 AI 代理开发。

The series includes MiniMax-Text-01, a foundation large language model (LLM), and MiniMax-VL-01, a visual multi-modal model.

该系列包括基础大语言模型 (LLM) MiniMax-Text-01 和视觉多模态模型 MiniMax-VL-01。

A massive context window

巨大的上下文窗口

MiniMax-Text-01, is of particular note for enabling up to 4 million tokens in its context window — the equivalent of a small library’s worth of books. The context window is how much information the LLM can handle in one input/output exchange, with words and concepts represented as numerical “tokens,” the LLM’s own internal mathematical abstraction of the data it was trained on.

MiniMax-Text-01 特别值得一提的是，它可以在其上下文窗口中启用多达 400 万个令牌，相当于一个小型图书馆的图书量。上下文窗口是法学硕士在一次输入/输出交换中可以处理的信息量，其中单词和概念表示为数字“标记”，这是法学硕士自己对其所训练的数据的内部数学抽象。

And, while Google previously led the pack with its Gemini 1.5 Pro model and 2 million token context window, MiniMax remarkably doubled that.

而且，虽然 Google 之前凭借其 Gemini 1.5 Pro 模型和 200 万个令牌上下文窗口处于领先地位，但 MiniMax 却将其显着增加了一倍。

As MiniMax posted on its official X account today: “MiniMax-01 efficiently processes up to 4M tokens — 20 to 32 times the capacity of other leading models. We believe MiniMax-01 is poised to support the anticipated surge in agent-related applications in the coming year, as agents increasingly require extended context handling capabilities and sustained memory.”

正如 MiniMax 今天在其官方 X 帐户上发布的那样：“MiniMax-01 可有效处理多达 400 万个代币，是其他领先型号容量的 20 至 32 倍。我们相信，随着代理越来越需要扩展的上下文处理功能和持续内存，MiniMax-01 已准备好支持来年预期激增的代理相关应用程序。”

The models are available now for download on Hugging Face and Github under a custom MiniMax license, for users to try directly on Hailuo AI Chat (a ChatGPT/Gemini/Claude competitor), and through MiniMax’s application programming interface (API), where third-party developers can link their own unique apps to them.

这些模型现在可以在定制的 MiniMax 许可下在 Hugging Face 和 Github 上下载，用户可以直接在 Hailuo AI Chat（ChatGPT/Gemini/Claude 的竞争对手）上进行尝试，并通过 MiniMax 的应用程序编程接口（API）进行尝试，其中第三方 -派对开发者可以将自己独特的应用程序链接到它们。

MiniMax is offering APIs for text and multi-modal processing at competitive rates:

MiniMax 以具有竞争力的价格提供用于文本和多模式处理的 API：

For comparison, OpenAI’s GPT-40 costs $2.50 per 1 million input tokens through its API, a staggering 12.5X more expensive.

相比之下，OpenAI 的 GPT-40 通过其 API 每 100 万个输入代币的成本为 2.5 美元，贵出惊人的 12.5 倍。

MiniMax has also integrated a mixture of experts (MoE) framework with 32 experts to optimize scalability. This design balances computational and memory efficiency while maintaining competitive performance on key benchmarks.

MiniMax 还集成了由 32 名专家组成的混合专家 (MoE) 框架，以优化可扩展性。这种设计平衡了计算和内存效率，同时在关键基准测试上保持了有竞争力的性能。

Striking new ground with Lightning Attention Architecture

利用闪电注意力架构开辟新天地

At the heart of MiniMax-01 is a Lightning Attention mechanism, an innovative alternative to transformer architecture.

MiniMax-01 的核心是闪电注意力机制，这是变压器架构的创新替代方案。

This design significantly reduces computational complexity. The models consist of 456 billion parameters, with 45.9 billion activated per inference.

这种设计显着降低了计算复杂度。这些模型由 4560 亿个参数组成，每次推理激活 459 亿个参数。

Unlike earlier architectures, Lightning Attention employs a mix of linear and traditional SoftMax layers, achieving near-linear complexity for long inputs. SoftMax, for those like myself who are new to the concept, are the transformation of input numerals into probabilities adding up to 1, so that the LLM can approximate which meaning of the input is likeliest.

与早期的架构不同，Lightning Attention 采用了线性和传统 SoftMax 层的混合，实现了长输入的近线性复杂性。对于像我这样刚接触这个概念的人来说，SoftMax 是将输入数字转换为加起来为 1 的概率，以便 LLM 可以近似输入最有可能的含义。

MiniMax has rebuilt its training and inference frameworks to support the Lightning Attention architecture. Key improvements include:

MiniMax 重建了其训练和推理框架以支持闪电注意力架构。主要改进包括：

These advancements make MiniMax-01 models accessible for real-world applications, while maintaining affordability.

这些进步使 MiniMax-01 模型可用于实际应用，同时保持经济实惠。

Performance and Benchmarks

性能和基准

On mainstream text and multi-modal benchmarks, MiniMax-01 rivals top-tier models like GPT-4 and Claude-3.5, with especially strong results on long-context evaluations. Notably, MiniMax-Text-01 achieved 100% accuracy on the Needle-In-A-Haystack task with a 4-million-token context.

在主流文本和多模态基准测试中，MiniMax-01 可以与 GPT-4 和 Claude-3.5 等顶级模型相媲美，尤其是在长上下文评估上取得了强劲的结果。值得注意的是，MiniMax-Text-01 在具有 400 万个令牌上下文的“大海捞针”任务中实现了 100% 的准确率。

The models also demonstrate minimal performance degradation as input length increases.

随着输入长度的增加，这些模型还表现出最小的性能下降。

MiniMax plans regular updates to expand the models’ capabilities, including code and multi-modal enhancements.

MiniMax 计划定期更新以扩展模型的功能，包括代码和多模式增强。

The company views open-sourcing as a step toward building foundational AI capabilities for the evolving AI agent landscape.

该公司将开源视为为不断发展的人工智能代理领域构建基础人工智能能力的一步。

With 2025 predicted to be a transformative year for AI agents, the need for sustained memory and efficient inter-agent communication is increasing. MiniMax’s innovations are designed to meet these challenges.

预计 2025 年将是 AI 智能体变革的一年，对持续记忆和高效智能体间通信的需求正在不断增加。 MiniMax 的创新旨在应对这些挑战。

Open to collaboration

开放合作

MiniMax invites developers and researchers to explore the capabilities of MiniMax-01. Beyond open-sourcing, its team welcomes technical suggestions and collaboration inquiries at model@minimaxi.com.

MiniMax 邀请开发人员和研究人员探索 MiniMax-01 的功能。除了开源之外，其团队还欢迎通过 model@minimaxi.com 提出技术建议和合作咨询。

With its commitment to cost-effective and scalable AI, MiniMax positions itself as a key player in shaping the AI agent era. The MiniMax-01 series offers an exciting opportunity for developers to push the boundaries of what long-context AI can achieve.

凭借对具有成本效益和可扩展的人工智能的承诺，MiniMax 将自己定位为塑造人工智能代理时代的关键参与者。 MiniMax-01 系列为开发人员提供了一个令人兴奋的机会，以突破长上下文 AI 的极限。