Do you want smarter insights in your inbox? Sign up for our weekly newsletters to get what is crucial for AI leaders, data and security. Subscribe now
Singaporek startup AI Manuswhich was on the first pages of newspapers at the beginning of this year for approaching a multi -level orchestration platform for consumers and “PRO” (professionals who want to conduct work operations) returns with an captivating recent operate of their technology.
While many other main competing AI suppliers, such as OpenAI, Google and XAI which have launched AI “Deep Research” or “Deep Resear” agents, who run minutes or hours of extensive internet research and write well digital, exact reports on behalf of users, Manus adopts a different approach.
Manus was previously reported as Claude anthropic models to feed its platform.
The AI Impact series returns to San Francisco – August 5
The next AI phase is here – are you ready? Join the leaders from Block, GSK and SAP to see the exclusive look at how autonomous agents transform the flows of the work of the company-decision-making in real time for comprehensive automation.
Secure your place now – the space is confined: https://bit.ly/3guplf
Parallel processing of research, summary and inventive exit
IN Video published on the official account XCo -founder Manus and Chief Scientist Yichao “Peak” Ji shows a demo of using wide research to compare 100 sneakers.
To perform the task, Manus Wide Research almost immediately rotates 100 parallel subagents – each assigned to the project analysis, prices and availability of one shoe.
The result is a sortable matrix supplied in both a spreadsheet format and a website within a few minutes.
The company suggests that wide studies are not confined to data analysis. It can also be used for inventive tasks, such as design exploration.
In one scenario, Manus agents generated poster designs in 50 separate visual styles, returning refined resources in the postal file.
According to Manus, this flexibility results from the system at the system level for parallel processing and communication of an agent-agent.
In the film, Peak explains that wide studies are the first application of optimized virtualization and agent architecture capable of calculated energy scaling 100 times outside the initial offers.
The function has been designed for automatic activation during tasks requiring immense -scale analysis, without the required manual switches or configuration.
Availability and prices
Wide research is available, starting from users Manus Pro plan and will gradually be available to people with plus and basic plans. For now, the prices of Manus subscription are structured as follows.
- Free – 0 USD/month includes 300 daily refresh loans, access to chat mode, 1 simultaneous task and 1 planned task.
- Basic – $ 19/month adds 1900 monthly loans (+1,900 bonuses during a confined offer), 2 simultaneous and 2 planned tasks, access to advanced models in agent mode, image generation/video/slides and exclusive data sources.
- Plus – an escalate of 39 USD/month to 3 parallel and 3 planned tasks, 3,900 monthly loans (+3 900 bonuses) and contains all the basic functions.
- Professional -199 USD/month offers 10 simultaneous and 10 planned tasks, 19,900 points (+19 900 bonuses), early access to beta function, Manus shirt and a full set of functions, including advanced agent tools and content generation.
There is also a 17% discount on these prices for users who want to pay in advance.
The launch is based on the infrastructure introduced with Manus at the beginning of this year, which the company describes not only an AI agent, but a personal cloud processing platform.
Each MANUS session works on a dedicated virtual machine, providing users with access to the organized cloud in a natural language in a natural language-which the company perceives as the key to enabling real flows of AI work.
Thanks to wide research, Manus users can delegate research or inventive exploration in dozens and even hundreds of subagents.
Unlike customary systems of many agents with predefined roles (such as a manager, coder or designer), each subagent in wide studies is a fully talented, fully distinguished manus instance-not specialized for a specific role-he does independently and can take any general task.
This architectural decision, according to the company, opens the door to the versatile, scalable task of unlimited service by unyielding templates.
What are the benefits of wide deep research?
It seems that the implication is that running all these agents in parallel is faster and will cause a better and more diverse set of work products outside of research reports, unlike individual “deep research” agents, which other AI suppliers showed or issued.
But although Manus promotes wide research as a breakthrough in the agent’s parallelism, the company does not provide direct evidence that the rebirth of dozens or hundreds of subagents is more effective than sequential tasks related to support for a single high -capacity agent.
The edition does not include reference points, comparisons or technical explanations in order to justify the compromises of this approach-as such as increased operate of resources, complexity of coordination or potential inefficiency. There is also a lack of details about the cooperation of subagents, how the results are integrated or whether the system offers measurable advantages of speed, accuracy or costs.
As a result, while the function presents architectural ambitions, its practical benefits in relation to simpler methods remain unverified depending on the information provided.
So far, the soles have mixed achievements so far …
While the implementation of wide research by Manus is set as progress in the general systems of AI agents, the wider ecosystem has recorded mixed results with similar subagative approaches.
For example, he Reddit, self -proclaimed Claude code users They aroused concerns about sluggish, consuming immense amounts of tokens and offering confined visibility in performance.
Common pain points include a lack of coordination protocols between agents, difficulties in debugging and irregular performance during high load periods.
These challenges do not necessarily reflect the implementation of Manus, but emphasize the complexity of the development of solid multi -stage frames.
Manus admits that wide studies are still experimental and can have some restrictions as it develops.
Looking to the future
With the introduction of wide research, Manus deepens its involvement in redefining the way users interaction with AI agents on a immense scale.
When other platforms are struggling with the technical challenges of coordination and reliability of the subagative, Manus’s approach can serve as a test case in the case of generalized agent-sea-resolved modules-can provide a vision of trouble-free, multiple cooperation of AI.
The company points to wider ambitions, suggesting that the infrastructure standing behind wide research is the basis for future offers. Both users and industry observers will pay special attention to whether this recent wave of agent architecture can meet their potential – or the challenges seen in a different place of space AI will eventually catch up.
Correction: In this article, it was originally incorrect that Manus was based in China when it is not; He is in Singapore. He also cited earlier reporting that the Alibaba QWEN models were used; That’s not. We updated and regretted mistakes.