Singaporean AI startup Manus, which gained attention earlier this year for its innovative multi-agent orchestration platform designed for both consumers and "pro"-sumers (professionals seeking to streamline work operations), has returned with a fascinating new application of its technology.

While major AI competitors like OpenAI, Google, and xAI have introduced “Deep Research” or “Deep Researcher” AI agents that perform extensive, in-depth web research and produce detailed, well-cited reports for users, Manus is opting for a different strategy.

The company has unveiled “Wide Research,” an experimental feature that allows users to carry out large-scale, high-volume tasks by utilizing parallelized AI agents — exceeding 100 agents at once, all working towards achieving a single task or a series of sub-tasks to accomplish an overarching objective.

Manus was previously noted for using Anthropic Claude models to power its platform.

AI Scaling Hits Its Limits

Power constraints, increasing token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to learn how leading teams are:

• Turning energy into a strategic advantage

• Designing efficient inference for real throughput gains

• Unlocking competitive ROI with sustainable AI systems

Secure your spot to stay ahead: https://bit.ly/4mwGngO

Parallel processing for research, summarization, and creative output

In a video posted on the official X account, Manus co-founder and Chief Scientist Yichao ‘Peak’ Ji demonstrates using Wide Research to evaluate 100 sneakers.

To accomplish the task, Manus Wide Research swiftly deploys 100 concurrent subagents — each tasked with examining one shoe’s design, pricing, and availability.

The outcome is a sortable matrix provided in both spreadsheet and webpage formats within minutes.

The company indicates that Wide Research extends beyond data analysis and can be applied to creative tasks such as design exploration.

In one instance, Manus agents simultaneously created poster designs in 50 distinct visual styles, delivering polished assets in a downloadable ZIP file.

According to Manus, this versatility is due to the system-level approach to parallel processing and agent-to-agent communication.

In the video, Peak explains that Wide Research is the inaugural application of an optimized virtualization and agent architecture capable of scaling computing power 100 times beyond initial capabilities.

The feature is designed to activate automatically during tasks that require wide-scale analysis, with no need for manual adjustments or configurations.

Availability and pricing

Wide Research is now available for users on the Manus Pro plan and will gradually become accessible to those on the Plus and Basic plans. Currently, Manus subscription pricing is structured monthly as follows.

• Free – $0/month Includes 300 daily refresh credits, access to Chat mode, 1 concurrent task, and 1 scheduled task.

• Basic – $19/month Adds 1,900 monthly credits (+1,900 bonus during limited offer), 2 concurrent and 2 scheduled tasks, access to advanced models in Agent mode, image/video/slides generation, and exclusive data sources.

• Plus – $39/month Increases to 3 concurrent and 3 scheduled tasks, 3,900 monthly credits (+3,900 bonus), and includes all Basic features.

• Pro – $199/month Offers 10 concurrent and 10 scheduled tasks, 19,900 credits (+19,900 bonus), early access to beta features, a Manus T-shirt, and the full feature set including advanced agent tools and content generation.

A 17% discount is also available for users who choose to pay annually upfront.

This launch builds on the infrastructure introduced with Manus earlier this year, which the company describes not just as an AI agent, but as a personal cloud computing platform.

Each Manus session operates on a dedicated virtual machine, granting users access to orchestrated cloud compute through natural language — a setup the company sees as essential for enabling true general-purpose AI workflows.

With Wide Research, Manus users can delegate research or creative exploration to dozens or even hundreds of subagents.

Unlike traditional multi-agent systems with predefined roles (such as manager, coder, or designer), each subagent within Wide Research is a fully capable, fully featured Manus instance — not limited to a specific role — operating independently and able to undertake any general task.

This architectural choice, the company asserts, allows for flexible, scalable task handling free from rigid templates.

What are the benefits of Wide over Deep Research?

The implication is that running these agents in parallel is faster and yields a more diverse set of work products beyond research reports, unlike the single “Deep Research” agents other AI providers offer or have implemented.

However, while Manus promotes Wide Research as a breakthrough in agent parallelism, the company does not offer direct evidence that deploying dozens or hundreds of subagents is more effective than a single, high-capacity agent managing tasks sequentially.

The release does not include performance benchmarks, comparisons, or technical justifications for the trade-offs of this method — such as increased resource usage, coordination complexity, or potential inefficiencies. It also lacks details on how subagents collaborate, how results are integrated, or whether the system provides measurable advantages in speed, accuracy, or cost.

Consequently, while the feature highlights architectural ambition, its practical benefits over simpler methods remain unproven based on the information provided.

Sub-agents have a mixed track record more generally, so far…

While Manus’s implementation of Wide Research is presented as a leap forward in general AI agent systems, the broader ecosystem has experienced mixed outcomes with similar subagent approaches.

For example, on Reddit, self-described users of Claude’s Code have expressed concerns about its subagents being slow, consuming large amounts of tokens, and providing limited visibility into execution.

Common issues include a lack of coordination protocols between agents, difficulties in debugging, and inconsistent performance during high-load periods.

These challenges don’t necessarily reflect on Manus’s implementation, but they underscore the complexity of developing robust multi-agent frameworks.

Manus acknowledges that Wide Research is still experimental and may have some limitations as development continues.

Looking ahead

With the introduction of Wide Research, Manus strengthens its commitment to redefining user interaction with AI agents at scale.

As other platforms grapple with the technical challenges of subagent coordination and reliability, Manus’s approach may serve as a test case for whether generalized agent instances — rather than narrowly focused modules — can achieve the vision of seamless, multi-threaded AI collaboration.

The company hints at broader ambitions, suggesting that the infrastructure supporting Wide Research sets the stage for future innovations. Users and industry observers will be closely watching to see if this new wave of agent architecture can fulfill its potential — or if the challenges observed elsewhere in the AI sector will eventually surface.

Correction: This article originally incorrectly stated that Manus was based in China when it is not; it is in Singapore. Also, it cited previous reports that it used Alibaba Qwen models; it does not. We have updated and regret the errors.