On August 2nd， as announced by 360 Group， NanoAI has officially completed its brand refresh and has been upgraded to “Multi-Agent Swarm.”
This signifies NanoAI’s transition to becoming the world’s first intelligent agent system to truly achieve L4 level capabilities. This advancement disrupts the traditional operational paradigm of intelligent agents， marking a significant evolutionary leap from “individual combat” to “swarm collaboration.” It represents a pivotal moment， propelling AI productivity into the “era of outcome delivery.”
From L1 to L4: The Four Evolutionary Stages of Intelligent Agents
Prior to the emergence of the NanoAI Multi-Agent Swarm， intelligent agents had progressed through three developmental phases:
—L1 Chat Assistants: Essentially conversational tools， adept at offering suggestions or emotional support， falling into the “toy-grade” category. Examples include GPTs， where AI operated in a “one-way response” mode， functioning more as passive information output “dialogue windows” and far from problem-solving at their core.
—L2 Low-Code Workflow Agents: Represented by workflow agents built with low-code methodologies， these evolved from “toys” to “tools.” However， they delivered semi-finished tools rather than final products.
—L3 Autonomous Planning Agents: Intelligent agents possessing reasoning capabilities， such as the initial version of Manus or the current 50，000+ specialized agents within NanoAI， including deep research agents and shopping agents.
While these agents demonstrated strong individual capabilities， akin to domain experts capable of “deep reasoning within a single domain” and multi-step logical deduction， even utilizing tools to achieve specific goals， they were limited by their technical frameworks when faced with complex cross-domain problems. Their lack of collaborative planning ability often led to bottlenecks， hindering global optimization through multi-layered thinking.
When the inherent limitations of single agents proved insufficient to overcome the hurdles of complex tasks， the innovative L4-level intelligent agent from NanoAI emerged. This upgrade transformed NanoAI’s product capabilities from “Super Intelligent Agents” to a “Multi-Agent Swarm.”
The Multi-Agent Swarm enables multiple intelligent agents to form teams and collaborate， much like assembling building blocks， by combining different specialized reasoning agents to serve users. At the L4 agent swarm level， it’s akin to hiring a team of professionals， with the flexibility to scale the team’s size at any time， achieving limitless expansion of capabilities and ultimately delivering results and achieving goals.
NanoAI Cracks the Technical Code for Multi-Agent Collaboration
While the AI landscape features numerous intelligent agents and explorations into “multi-agent” systems， significant technical challenges remain.
Practical tests have shown that current popular multi-agent technical frameworks struggle to address issues such as task allocation， parameter passing， and context management， which are crucial for effective team collaboration. If a single agent has a 90% success rate， the success rate for five collaborating agents can drop below 50%.
In multi-step workflows， error rates compound exponentially. With a 95% reliability at each step， after 20 steps， the overall success rate dwindles to approximately 30%. Without resolving these issues， multi-agent collaboration risks falling into a “too many cooks spoil the broth” scenario， where the combined effect is far less than the sum of its parts.
To overcome this， NanoAI has pioneered a novel “Swarm Collaboration Framework.” This framework allows multiple reasoning agents to flexibly form teams， nest within layers， and collaborate to complete complex tasks. This technology leads the industry in two key aspects:
Firstly， it enables multiple agents to flexibly form teams， nest in layers， and collaborate on complex tasks， fostering tight cooperation akin to that of a bee swarm. This is not about a single agent working in isolation， but rather multiple highly capable “AI workers” phân công công việc， each fulfilling their role concurrently， leading to a surge in efficiency and outcome delivery.
Secondly， the upgraded NanoAI successfully implements flexible configuration of the “Multi-Agent Swarm，” allowing for both single-swarm operations and the formation of swarm formations. In response to new user tasks， these formations can be dynamically reconfigured， embodying the principle of “coming together as a fiery force， and scattering as a thousand stars.”
Essentially， NanoAI’s L4 swarm acts as a collaborative hub for L3 experts， ensuring that specialized agents interlock and coordinate like precision gears. It also functions as a collaborative network for super-clusters， connecting multiple swarms to synergize their efforts. From deep specialization in single domains to cross-system collaboration， each team formation expands the boundaries of capability， and each layer of aggregation drives more efficient outcome delivery.
As of now， NanoAI hosts over 50，000 L3 reasoning agents， each possessing formidable individual combat capabilities. Within NanoAI， users can design and build reasoning agents using natural language， democratizing the creation of personal “Manus.”
Under the swarm framework， these agents are organized into L4 teams that execute tasks with unified objectives and synchronized steps under a single command system. This ensures that the role of each agent within the team is maximized， ultimately completing highly complex and extended tasks， such as producing a movie-quality blockbuster lasting 10 minutes.
Currently， the NanoAI Agent Swarm can execute over 1000 steps for tasks uninterrupted for 2 continuous hours， consuming over 20 million tokens.
The secret to the formidable capabilities of the NanoAI Multi-Agent Swarm lies in its industry-leading multi-agent swarm engine. This engine is the core that empowers the swarm to support unlimited tool invocation， unlimited context length， and execute super-tasks of infinite steps.
NanoAI has launched over 10 categories of multi-agent swarms， covering scenarios such as video production， content creation， industry research， e-commerce live streaming， and travel planning. These swarms are customized and led by industry experts， offering a more informed and practical experience.
NanoAI Multi-Agent Swarm Achieves “Generate Blockbusters with a Single Sentence” Capability
NanoAI has developed China’s first “Generate Blockbusters with a Single Sentence” Multi-Agent Swarm. For tasks like scriptwriting， storyboarding， visual creation， voice-over， music composition， and final editing， using L1 to L3 level agents would typically take at least two hours. Now， it can be accomplished in just 20 minutes.
Hit Video Generation Multi-Agent Swarm: This is your AI-powered “MCN Team.” From hot topic monitoring and script generation to video creation， you can effortlessly launch your own self-media account and create trending short videos with a single prompt.
E-commerce Live Streaming Video Multi-Agent Swarm: This swarm can instantly translate your live streaming videos into multiple languages. You can also upload a photo of yourself along with a product image to generate a promotional video on the spot.
Just as the Industrial Revolution powered the world with steam and the internet shattered geographical information barriers， the core of every productivity revolution lies in the confrontation with previously unimaginable possibilities beyond the “known.”
These advancements challenge the very boundaries of human imagination regarding “possibility.” NanoAI’s “Multi-Agent Swarm” continues this narrative， extending beyond efficient execution of known tasks to breaking through the ceiling of “what humans can conceive.”
The concepts that were once shelved due to the need for cross-domain teams， long execution times， and coordination difficulties， along with innovations unattainable by individual experience and skills， are now being unlocked by swarm-like intelligent collaboration.
AI’s role is precisely to help humanity transcend the limitations of cognition and capability， transforming “what cannot be imagined” into the starting point for “what can be achieved.” The ultimate significance of this revolution is not to replace human hands， but to provide humanity’s imagination with unprecedented reach and extension.
