Infini AI Launches Enterprise Agent Platform for Creativity

[BODY] Infini AI has introduced an agent service platform designed to enhance enterprise creativity and productivity through intelligent agents. The company asserts that future enterprises will be "Agentic," driven by hundreds or thousands of intelligent agents that can amplify creative output.

Xia Lixue, co-founder and CEO of Infini AI, stated that an organization's creative capacity would be "infinitely amplified" with 10 to 1,000 agents active within an enterprise, enabling even small teams to achieve significant impact. The Infini AI Agent Service Platform was officially unveiled at the 2025 INFINI DAY Infini AI Agent Ecosystem Forum on December 16. The platform aims to provide full-lifecycle services for enterprises, including agent customization, optimization, deployment, hosting, commercialization, and monetization.

Addressing Agent Deployment Challenges

Infini AI posits that while agentic AI accelerates the scaling of enterprise creativity, the high threshold for production-grade agent deployment and scaling presents a significant obstacle. Li Boxun, CTO of Infini AI, identified four core challenges for enterprises in agent deployment: difficulty in achieving production-grade results, ensuring stable operation, controlling costs, and enabling commercialization.

Li emphasized that resolving these issues and bridging the gap between infrastructure and agent development is crucial for enterprise-grade agents to evolve from creative tools to productive assets.

Platform Capabilities and Features

The Infini AI Agent Service Platform offers out-of-the-box agent capability templates and reliable hosting services, integrated with Infini AI's underlying computing power and model infrastructure. It provides full-lifecycle support, from agent customization to commercialization, aiming to make agent production more effective, operations more stable, costs more controllable, and commercialization smoother.

The platform incorporates industry knowledge and over five types of agent capability templates, including code, research, and multimodal functions, to help agents achieve initial results. Infini AI has templated industry experience from large customers, making it reusable for small and medium-sized enterprises (SMEs). The platform also tracks cutting-edge models and dynamically adapts solutions for target business effects. Customization options, such as tool calls and prompt optimization, are available to improve agent application for specific enterprise needs.

For stability and reliability, the platform leverages Infini AI's computing power reserves, exceeding 25,000P, and infrastructure support. It features elastic scaling and automated troubleshooting, with thousand-card cluster model service stability reaching 99.5%. Sandbox scheduling speed has been reduced from hundreds of milliseconds to tens of milliseconds. An observable dashboard allows enterprises to track agent lifecycle data, ensuring secure operation and iterative expansion.

In terms of cost control, the platform integrates with underlying infrastructure, offering model integration resources, inference optimization technology, and full-stack software and hardware collaborative capabilities. It gathers and updates over 20 mainstream and cutting-edge large models. For models with hundreds of billions or trillions of parameters, such as DeepSeek and Kimi-K2-Instruct, the platform uses a large-scale PD separation solution, improving inference efficiency by two to three times compared to traditional service models. It also supports model customization and system-level tuning to optimize performance and cost for specific application scenarios.

The platform addresses efficiency chain breaks in commercial deployment and scaled service of intelligent agents by supporting access to community or self-developed toolsets, reducing integration effort by over 70%. It encapsulates complex agent business processes into independent modules for version management and updates. Additionally, it provides quick access to third-party marketing and payment channels, such as WeChat and Xiaohongshu, to bridge commercialization gaps.

Real-World Application

Infini AI showcased the "SysCoding Agent," an enterprise system development agent developed in collaboration with an operations and maintenance service enterprise. This agent, built on the Infini AI Agent Service Platform, allows employees, partners, and customers to create, modify, and host business systems through natural language dialogue, integrating with enterprise resources and tool libraries.

The SysCoding Agent reportedly achieves over 95% completeness in its main process, exceeds 90% compliance with specifications, and has a blocking bug incidence of less than 3% in its first generation. This means users can obtain a usable version meeting most requirements through initial interaction. The agent is currently in internal use, handling peak work period concurrent requests. One person can complete the development and launch of a production-grade system in one week, with a single system construction cost as low as 5 yuan. Similar productivity gains are observed in industries such as job searching, tourism, and education.

Dai Guohao, Associate Professor at Shanghai Jiao Tong University and co-founder and Chief Scientist of Infini AI, stated that Infini AI will continue to support the evolution of agent application efficiency in digital and physical worlds through "Agentic Infra" and industry-academia-research collaboration, building an innovation chain for agents.