DeepSeek is expanding its ambitions beyond artificial intelligence models and chatbots. Its open-source AI agent platform, DeepSeek Harness, reportedly reached 1.54 million daily active users by November 4, less than three months after its debut in August. The milestone highlights the growing demand for AI systems that can perform multi-step tasks, interact with software tools and complete work with less human supervision.
The Chinese AI company is also developing an official plugin marketplace and a support programme for developers who build extensions for Harness. The platform reportedly processes approximately 92 trillion tokens per day and has attracted around 35,000 third-party plugins. These developments suggest DeepSeek wants to compete with OpenAI and Anthropic not only through its AI models but also through the software ecosystem built around them.
Key Takeaways
- DeepSeek Harness reportedly reached 1.54 million daily active users by November 4, 2026.
- The platform launched on August 13, 2026, and introduced a desktop version in September.
- Daily active users increased from approximately 271,000 at the end of September.
- DeepSeek says Harness processes around 92 trillion tokens daily.
- The platform has approximately 35,000 third-party plugins.
- Around 60% of users have reportedly created or installed at least one third-party plugin.
- DeepSeek plans an official plugin marketplace and a support programme for plugin developers.
- The company is positioning Harness as an open agent ecosystem to compete with established AI platforms.
DeepSeek Harness Reaches 1.54 Million Daily Active Users
DeepSeek disclosed that its open-source agent application, Harness, had reached approximately 1.54 million daily active users by November 4. The milestone came less than three months after the platform’s initial release on August 13.
The growth has been particularly notable since the introduction of its desktop version. DeepSeek’s reported figures show that daily users stood at approximately 271,000 around the end of September. The November figure represents a substantial increase in the number of users connecting to the company’s infrastructure.
The figures were disclosed by DeepSeek and reported by Bloomberg. They are based on connections to DeepSeek’s own servers and should not be interpreted as an independently audited measurement of every person using the software.
The company is releasing these adoption numbers as competition in artificial intelligence increasingly moves beyond traditional conversational interfaces. Rather than simply answering questions, AI agents are designed to carry out sequences of actions, use external tools and help users complete more complicated tasks.
For DeepSeek, growing adoption of Harness could provide an opportunity to establish a position in this emerging category.
What Is DeepSeek Harness?
DeepSeek Harness is an open-source framework designed to help developers and users build and operate AI agents.
A conventional chatbot generally responds to a user’s prompt with generated text, code or other information. An AI agent can take a more active role by breaking a request into smaller tasks, selecting tools, interacting with files and applications, and continuing until it reaches a desired result.
For example, a user could ask an agent to examine a software project, identify a programming problem, modify the relevant files and run tests. Instead of providing only instructions, the agent can use available tools to carry out parts of the workflow.
Harness provides the infrastructure for these activities. It is built around an architecture in which different capabilities can be added, replaced or extended through plugins.
DeepSeek describes this approach as an everything-is-a-plugin architecture. Its components can include models, tools, skills, sessions, storage, sandboxes and user-interface elements.
This modular design gives developers greater flexibility than a system in which every component is fixed by the platform provider. Developers can customise the environment for different applications and workflows.
The official project is available as open-source software under the MIT licence. However, its repository describes the product as being in developer preview and warns that compatibility-breaking changes may occur as development continues.
DeepSeek’s 92 Trillion Daily Tokens Highlight the Computing Challenge
One of the most significant figures associated with Harness is its reported daily token consumption.
DeepSeek says the platform processes approximately 92 trillion tokens every day. Based on 1.54 million daily active users, that works out to roughly 60 million tokens per daily user.
A token is a basic unit of text or code processed by an AI model. Token consumption is an important measure because AI companies use it to track how much their models are being used and, in many cases, to calculate computing costs.
The figure illustrates why AI agents could create a different infrastructure challenge from traditional chatbots.
A typical chatbot interaction might involve a user asking a question and receiving a response. An agent may instead work through dozens or hundreds of intermediate steps, depending on the task. It could inspect documents, write code, execute commands, search the internet, call external tools and review its own output.
These operations can generate substantially more model activity than a single question-and-answer exchange.
| Metric | Reported figure |
|---|---|
| Daily active users | 1.54 million |
| Daily token consumption | 92 trillion |
| Approximate tokens per daily user | 60 million |
| Third-party plugins | 35,000 |
| Users creating or installing plugins | Around 60% |
The per-user token figure is an average calculated from the reported totals. It does not mean every user consumes the same amount. A relatively small number of intensive users could account for a large share of total processing.
The numbers also have implications for the wider AI infrastructure market. If agent-based applications become more common, demand could rise for computing capacity, memory, networking and data-centre infrastructure.
However, high token consumption does not automatically translate into profitability. The economics will depend on model efficiency, hardware costs, pricing and how frequently users return to the platform.
DeepSeek Plans an Official Plugin Marketplace
DeepSeek is preparing an official marketplace where developers can publish plugins and users can discover additional capabilities for Harness.
Plugins are software extensions that add specific functions to an existing application. They can allow an AI agent to interact with services, perform specialised operations or support workflows that are not included in its basic installation.
The marketplace could make it easier for users to find and install tools without having to locate individual projects across developer communities.
DeepSeek has reportedly counted approximately 35,000 third-party plugins. The company also says around 60% of users have created or installed at least one plugin, although the quality and technical sophistication of these extensions vary.
The planned marketplace could help organise this growing collection and give developers a more visible distribution channel.
An official marketplace may also help DeepSeek strengthen the relationship between the platform and its developer community. If users can easily discover useful extensions, they may have more reasons to continue using Harness.
Planned support for plugin developers
DeepSeek is also preparing a support programme for plugin creators. The proposed assistance includes:
- Token incentives: AI usage credits that developers can use while building and testing plugins.
- Technical assistance: Support to help developers improve their extensions and integrate them with Harness.
- Promotion: Opportunities to increase the visibility of useful plugins and their creators.
- Developer engagement: Greater collaboration between DeepSeek and the people building tools for its ecosystem.
The company has not disclosed a confirmed launch date, detailed eligibility requirements or the financial terms of the programme.
The success of this initiative will depend on whether DeepSeek can attract developers who create reliable, useful and regularly maintained extensions.
DeepSeek Takes on OpenAI and Anthropic in the AI Agent Market
The growth of Harness comes as leading AI companies compete to build products that can perform work rather than simply generate responses.
OpenAI has expanded its agent-related offerings through products such as Codex, while Anthropic has developed Claude Code and other tools for software development and workplace tasks.
These companies are competing across several layers of the AI market, including model performance, developer tools, integrations and the ability to complete complicated workflows.
DeepSeek’s approach is different in an important respect: Harness is open source and built around modular components that developers can customise.
That could appeal to developers who want greater control over the software powering their agents. It could also encourage experimentation because users can extend the platform rather than depending entirely on features provided by a single company.
However, open-source availability alone does not guarantee that Harness will become a leading enterprise platform. Reliability, security, documentation, compatibility and long-term maintenance will all influence adoption.
DeepSeek will also need to demonstrate that its ecosystem can support demanding real-world applications, rather than relying only on headline user numbers.
Why AI Agents Could Become the Next Major Technology Market
The AI industry is gradually moving from conversational systems towards software capable of executing tasks.
The first major wave of generative AI adoption focused on chatbots that could write, summarise, translate, explain and generate code. The next phase could involve agents that connect these capabilities to external software and complete workflows.
Consider a business that needs to analyse a collection of documents, extract financial information, prepare a report and organise the findings into a spreadsheet. A traditional chatbot can help with individual steps, but an agent could potentially coordinate several operations through connected tools.
The same principle applies to software development, research, customer support and administrative work.
This creates a new competitive opportunity. AI companies may increasingly compete over which platform has the most useful tools, integrations and developer-created extensions.
For DeepSeek, Harness provides a way to participate in that market without relying exclusively on the popularity of its flagship models.
The plugin strategy could become particularly important because specialised tools make an agent more useful for specific industries and tasks. A strong developer ecosystem can also create network effects: more developers produce more extensions, which attract more users, potentially encouraging still more development.
Challenges Facing DeepSeek Harness
Despite its reported growth, Harness faces several challenges.
Plugin quality and security: A large plugin ecosystem can introduce inconsistent quality, compatibility problems and security risks. DeepSeek will need effective ways to identify malicious or unreliable extensions.
Platform reliability: AI agents often perform multiple operations in sequence. A failure in one step can interrupt an entire workflow, making reliability especially important.
Computing costs: Processing approximately 92 trillion tokens daily could require substantial computing resources. DeepSeek must find ways to manage infrastructure costs as usage grows.
Developer retention: The planned marketplace and support programme could attract developers, but long-term participation will depend on the quality of documentation, tools and incentives.
Competition: OpenAI and Anthropic already have established developer products and commercial ecosystems. DeepSeek will need to offer compelling advantages to win users and businesses.
There is also an important distinction between adoption and commercial success. Daily active users and token consumption show activity, but they do not reveal revenue, profitability, customer retention or the cost of serving each user.
The Bigger Picture
DeepSeek’s reported 1.54 million daily active users suggest that the company is making progress in a part of the AI industry that could become increasingly important. The competition is no longer limited to building more capable language models. Companies are also trying to control the infrastructure through which those models interact with applications, data and external services.
Harness gives DeepSeek an opportunity to compete at this application and infrastructure layer. Its open-source architecture, growing plugin collection and planned developer marketplace could help it build an ecosystem around its AI technology.
The reported 92 trillion daily tokens also highlight a challenge for the industry. More capable agents may generate significantly greater computing demand because they perform extended sequences of operations. This could create opportunities for semiconductor manufacturers, cloud providers and data-centre operators, while putting pressure on AI companies to improve efficiency.
Frequently Asked Questions
1. What is DeepSeek Harness?
DeepSeek Harness is an open-source AI agent framework developed by DeepSeek. It allows users and developers to build and operate agents that can use tools, work with files, conduct research and perform coding-related tasks.
2. How many daily users does DeepSeek Harness have?
DeepSeek reported approximately 1.54 million daily active users by November 4, 2026. The figure was reported by Bloomberg and is based on users connecting to DeepSeek’s own servers.
3. How many tokens does DeepSeek Harness process daily?
DeepSeek says Harness processes approximately 92 trillion tokens per day. This works out to roughly 60 million tokens per daily active user when calculated using the reported user count.
4. How many plugins does DeepSeek Harness have?
DeepSeek reportedly counts approximately 35,000 third-party plugins. Around 60% of users have created or installed at least one plugin, according to the company’s figures.
5. Is DeepSeek launching a plugin marketplace?
Yes. DeepSeek is planning an official marketplace where developers can publish plugins and users can discover them. The company has not announced a confirmed launch date.
6. How is DeepSeek Harness different from a chatbot?
A chatbot primarily responds to prompts, while an AI agent can use tools and perform multiple connected actions to complete a task. Harness provides an environment for building and operating these agents.
7. Is DeepSeek Harness open source?
Yes. DeepSeek makes Harness available as open-source software under the MIT licence. Its official repository currently identifies the project as being in developer preview.
Looking Ahead
DeepSeek’s next challenge will be converting its early user growth into a dependable developer ecosystem. The planned plugin marketplace and creator-support programme could help the company attract more developers, improve the availability of specialised tools and make Harness useful across a wider range of workflows. The quality, safety and reliability of those extensions will be just as important as their total number.
The wider AI industry will be watching whether open-source agent infrastructure can compete with the integrated products being developed by OpenAI and Anthropic. If DeepSeek can maintain user growth while controlling computing costs and improving its platform, Harness could become an important part of its strategy to expand beyond AI models and into the next generation of software.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
