DeepSeek’s use agent framework reached 1.54 million daily active users less than three months after launch, according to figures reported by the South China Morning Post on November 9. The Hangzhou-based company said users accessing use through its official API were consuming about 92 trillion tokens a day as of November 4. The numbers indicate that DeepSeek is seeking a larger role in the software infrastructure around AI agents, not only in the foundation models that power them. The user and token figures are company-reported metrics and have not been independently audited in the cited report.
use is a software layer for connecting models with tools and execution
A use is the software layer that connects a language model to tools, memory, sessions, code execution, scheduling and other capabilities needed to complete multi-step tasks. A basic chatbot can return a response to a prompt, but an agent needs to plan actions, call tools, inspect results and decide what to do next. The use manages those interactions and provides the runtime environment in which the model operates. As a result, the framework can influence reliability, security and cost even when the underlying model remains unchanged.
DeepSeek’s reported token consumption is large, but the figure should not be treated as a direct measure of unique human users or successful tasks. Tokens can be generated by repeated tool calls, long contexts, automated workflows and agent loops. One active user may trigger many requests, while many users may generate relatively little traffic. To understand the platform’s commercial significance, observers would need information about retention, paid usage, task completion, developer adoption and the proportion of consumption that is production traffic rather than testing.
Adoption figures are company-reported and need context
The move into agent infrastructure also changes the competitive landscape. Model providers have incentives to provide a complete development environment because developers who build workflows around a particular use may become more closely tied to its APIs, tools and deployment conventions. An open or modular design can attract users who want to switch models, but it still needs stable interfaces, permission controls and reliable execution. DeepSeek will need to show that use can handle credentials safely, isolate untrusted code, recover from failures and provide observability when agents execute tasks over long periods.
The adoption figures suggest that DeepSeek is testing a route to monetisation beyond model access. Agent platforms can generate recurring demand for inference, tool integrations and managed execution. They can also create higher switching costs if teams store workflow definitions, memory and operational data in one ecosystem. Developers evaluating use should test how easily they can replace the underlying model, export state, restrict tools and audit agent actions. The headline usage figures are notable, but durable success will depend on whether the framework becomes dependable infrastructure for real applications rather than a short-lived surge in developer experimentation.
The platform’s growth will be easier to evaluate when DeepSeek publishes more detail about the composition of its user base. Daily active users can include developers testing an API, companies running production workflows and automated clients that interact with the service repeatedly. Those groups have different commercial value and support requirements. Token consumption can indicate workload intensity, but it does not reveal whether agents complete tasks correctly or whether customers are willing to pay for the service. Useful future metrics would include retention, paid conversion, task success, average cost per completed workflow and the number of developers deploying production applications. Those measures would show whether use is becoming a durable platform rather than attracting short-term experimentation.
DeepSeek’s next challenge is to convert interest into trust among developers who deploy agents with access to business data. Clear documentation, stable APIs, transparent pricing and reliable support will matter as much as raw token throughput. Without those foundations, a large usage figure may not translate into long-term enterprise adoption.