Skip to content
Main Site News Console

3x Faster Than Claude Code with Truly Infinite Context: APUS AI Open-Sources Reflex Agent

· 量子位
国内AI

On October 13, APUS AI (Qilin Hesheng Network Technology Co., Ltd.) announced the open-sourcing of its Personal Agent product, Reflex Agent. Built with a harness powered by its self-developed, open-source decision model Reflex, and integrated with the Continuum memory engine, the product drives core harness control logic directly through the model—enabling higher decision-making efficiency and long-horizon task execution.

Image 3

According to benchmarks by the APUS AI Lab, Reflex Agent achieves response speeds up to 3x faster than mainstream agents such as Claude Code and Codex in relevant evaluation scenarios.

Unlike traditional architectures that rely primarily on hardcoded logic and prompts to orchestrate the harness, Reflex Agent delegates key decision-making processes—including model routing, skill selection, result review, and execution supervision—directly to the Reflex decision model. It drives tasks forward through a closed-loop mechanism of “decision—execution—feedback—re-decision.” APUS AI refers to this mechanism as “Loop Working,” underscoring that agents should not be confined to passively answering questions, but should continuously and autonomously take action aligned with user goals.

In terms of memory capabilities, Reflex Agent deeply integrates the Continuum memory engine, featuring three core characteristics:

  • Infinite Memory: Continuously accumulates and preserves end-to-end task information;
  • Lossless Memory: Fully preserves original content alongside its contextual relationships;
  • Lossless Input: Ensures relevant historical information feeds into subsequent decisions in its native form whenever possible.

This allows the agent to retrieve and invoke historical context on demand throughout long-horizon, complex tasks, drastically reducing reliance on information compression and summarization, and providing the underlying foundation for true “infinite context.” It is worth noting that “infinite context” here does not refer to bypassing the physical context window limit of a single model call; rather, it ensures that an agent’s memory capability and task lifecycle are no longer bound to a single context window.

This open-source release marks a key step in APUS AI’s exploration of the Loop Working paradigm and the development of next-generation Personal Agents. Reflex Agent is open-sourced under the MIT license, offering desktop, web, and SDK interfaces for developers to experience, reproduce, and build upon.

#国内AI#Claude#AI智能体#开源#4Allapi.com

Published by the 4All API team

Original link:https://www.qbitai.com/2026/10/503165.html

Need a mainstream LLM API? 4All API gives you one key to call OpenAI, Anthropic, Google Gemini, Qwen, DeepSeek, and dozens more — at official-pass-through pricing, with enterprise-grade reliability, integrated in 5 minutes.

Sign up for the 4All API console →