MemoraX AI Closes Multi-Million Dollar Seed Round

Memory-Focused Large Model Sector Attracts Fresh Capital

MemoraX AI, a startup specializing in memory-centric large models, has officially announced the completion of a multi-million dollar seed round, co-led by L2F Lightsource Founders Fund and Zhongding Capital. The proceeds will primarily fund the iteration and engineering deployment of its Agentic RL (agent-based reinforcement learning) algorithms, as well as the productization of its endogenous memory module.

Targeting the Core Capability of 'Memory'

MemoraX AI is tackling a critical yet long-overlooked challenge in the large model space — memory. While today's mainstream large models excel in reasoning and generation, they still exhibit significant weaknesses in long-term memory management and context persistence. AI agents frequently struggle with failing to remember or inaccurately recalling information when executing complex tasks, severely limiting their usability in real-world business scenarios.

MemoraX AI is dedicated to building an "endogenous memory module" for large models, enabling AI systems to efficiently store, retrieve, and update information across multi-turn interactions and long-horizon tasks. This approach aims to dramatically improve the continuity and consistency of AI agents.

Technical Approach: Agentic RL Driving Memory Evolution

Notably, MemoraX AI has adopted Agentic RL (agent-based reinforcement learning) as its core technical approach. Unlike traditional memory-augmented solutions, Agentic RL enables agents to autonomously learn "what to remember," "when to remember," and "how to recall memories" through continuous interaction with their environment, achieving dynamic evolution of memory capabilities.

This technical direction aligns precisely with the broader industry shift from "single-pass inference" to "persistent AI agents." As AI Agent use cases expand rapidly, agents equipped with robust memory capabilities are poised to deliver greater value in domains such as personal assistants, enterprise automation, and research support.

Investors Bet on Long-Term Value

The co-lead investors in this round, L2F Lightsource Founders Fund and Zhongding Capital, are both well-known investment institutions in China. L2F Lightsource Founders Fund has a long-standing focus on cutting-edge technology, while Zhongding Capital boasts deep investment experience in supply chain and enterprise services. Their joint lead investment reflects both confidence in MemoraX AI's technical capabilities and a strategic conviction in the memory-focused large model segment.

Industry Outlook

As competition in the large model space gradually shifts from "general capability benchmarks" to "deep vertical specialization," memory is emerging as one of the core differentiators for next-generation AI systems. MemoraX AI's latest funding round injects fresh momentum into this nascent sector.

Looking ahead, as Agentic RL algorithms continue to mature and the endogenous memory module moves toward productization, MemoraX AI is well-positioned to secure a significant role in the AI Agent infrastructure layer, providing critical memory capabilities for intelligent agent applications across industries.