Today, we’re excited to officially introduce MiniMax Agent, a general intelligent agent designed to tackle long-horizon, complex tasks. It can perform expert-level multi-step planning, flexibly break down task requirements, and execute multiple subtasks to deliver complete end-to-end solutions.
MiniMax Agent has already been used internally for nearly 60 days and has become a daily tool for over 50% of our team.
Just two days ago, we released our latest MiniMax-M1 model. You can now ask MiniMax Agent to generate a 15-minute quick-read overview to help you grasp it faster:
https://11zrgrdvcp.space.minimax.io/
Of course, it can also create learning tutorials with audio narration:
https://8t97qxgrm2.space.minimax.io/
To make up for not watching EVA as a kid, here’s a quick crash course — all generated by MiniMax Agent:
https://01pr7piao3.space.minimax.io/
Beyond that, our developers have used MiniMax Agent to experiment with advanced front-end animations ahead of our product launch:
https://kkyd1s0gol.space.minimax.io/
It can even generate complete product pages, for example, built an online Louvre in 3 minutes:
https://ze359rpe43.space.minimax.io/
Our journey building MiniMax Agent
About six months ago, we started with a vertical-specific agent called "Today's Personalized News", designed to follow custom information streams through defined workflows. We quickly realized this was just a subset of what a flexible, general-purpose agent could do — so we aimed higher, developing a truly general agent with much greater potential.
During development, we designed and evaluated the Agent by the standards of a "reliable teammate."
1. Programming capabilities
a. Handles complex logic with multiple components and branching flows
b. Simulates user operations to perform end-to-end testing — it won’t ship buggy pages
c. Prioritizes visual design and interaction quality for a great UX/UI
2. Multimodal understanding & generation
a. Understands long-form text, videos, audio, and images
b. Supports generation of images, audio, and video, enabling rich multimedia output from a single prompt
3. MCP integration
a. Natively integrated with MiniMax MCP for powerful, cost-effective multimodal output
b. Supports common tools like GitHub/GitLab, Slack, Figma, etc., extending context for productivity and creativity
Currently, MiniMax Agent relies on multiple models rather than a unified one, which introduces some overhead in cost and efficiency. We’re actively working to make it more efficient and affordable, so it can be a practical tool for everyone's everyday use.
Here’s a classic saying:
“Talk is cheap, show me the code.”
But when working with an intelligent agent, you start to feel a shift — not just in productivity, but in how work is structured.
Maybe it’s time to say:
“Code is cheap，show me the requirement.”
Last month, IMF projected that AI could add 0.5% to global GDP growth annually between 2025 and 2030. Whether we surpass that projection — only time will tell.
Intelligence with Everyone.
