---
type: "raw_signal"
id: "qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab"
slug: "qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab"
title: "Qwen released QwenGyre , an elastic RL framework for training agents on hours-long tasks."
description: "## QwenGyre targets the training bottlenecks of long-horizon agents"
retrieval_nugget: "## QwenGyre targets the training bottlenecks of long-horizon agents"
observed_at: "2026-10-12"
record_date: "2026-10-12"
date_kind: "observed_at"
why_it_matters: "Source: https://arxiv.org/abs/2609.33848"
novelty: "notable"
verification_level: "source-linked"
signal_type: "research"
source_platform: "arxiv.org"
topics: []
entities: ["Qwen","QwenGyre","Qwen3.8 2.4T","NL2RepoBench","Colocate","Async"]
related_patterns: ["agent-economics-moves-to-completed-work","harness-architecture-outlives-model-choice"]
source_url: "https://arxiv.org/abs/2609.33848"
source_urls: ["https://arxiv.org/abs/2609.33848"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/signals/qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab/","markdown":"https://newruntime.com/signals/qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab.md","json":"https://newruntime.com/signals/qwen-released-qwengyre-an-elastic-rl-framework-for-0667ab.json"}
---

# Qwen released QwenGyre , an elastic RL framework for training agents on hours-long tasks.

## Retrieval answer

## QwenGyre targets the training bottlenecks of long-horizon agents

## QwenGyre targets the training bottlenecks of long-horizon agents

A Qwen research paper introduces QwenGyre, an end-to-end framework for online reinforcement learning on extreme-long-horizon agent tasks. The paper describes executions lasting hours, spanning hundreds of model-environment interactions and approaching one million tokens per rollout. Its central systems mechanism is elastic GPU allocation between rollout generation and training without interrupting live executions. A trajectory processor reconstructs branching histories, scores partial progress, and deduplicates redundant paths.

At scale with Qwen3.8 2.4T and 700,000-token rollouts, the authors report a 6.0 percentage-point improvement on NL2RepoBench, from 52.5% to 58.5%, in 48 steps. Across evaluations on diverse training-data domains, they report speedups of up to 1.85× over Colocate and 1.78× over Async.

These are results reported in the paper, not an independent replication. The supplied arXiv record does not establish implementation availability. The work’s immediate contribution is a training architecture aimed at reducing GPU idle time and redundant trajectory cost under long, variable executions.

Source: https://arxiv.org/abs/2609.33848

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)