# Microsoft releases Agent Lightning rebuild for training existing agent systems

_Published Wednesday, October 7, 2026 at 10:35 PM EDT · AI · Latest · Tier 2 — Notable_

![Microsoft releases Agent Lightning rebuild for training existing agent systems — Primary](https://www.microsoft.com/en-us/research/wp-content/uploads/2026/10/AgentLightning-TWLIFB-1200x627-1.jpg)

Microsoft Research Asia has open-sourced Agent Lightning v1.0, a rebuilt framework that trains AI agents through the same software that coordinates their deployed work. An intermediary model endpoint records requests and responses, allowing existing agent code to remain unchanged.

Agents run as standard Kubernetes jobs or local processes without paid commercial sandbox services. Microsoft reports that its coding pipeline raised Qwen3.5-9B's SWE-bench Verified Pass@1 score from 41.8% to 56.4%, using about 6,000 training samples.

The framework also lets task execution and model updates share GPUs. In experiments, that approach achieved about a 2x end-to-end speedup over synchronous training, rather than waiting for the slowest agent in each batch.

## Sources

- [Microsoft Research](https://www.microsoft.com/en-us/research/blog/agent-lightning-v1-0-a-3500-line-lightweight-agentic-rl-framework-for-training-agents-with-real-harnesses/)

---
Canonical: https://techandbusiness.org/newswire/T9dJMsatZ0-FxOSI4Z7wrO
Published: 2026-10-08T02:35:12.602Z
Story chronology: 2026-10-07T16:00:00.000Z
Retrieved: 2026-10-08T04:44:12.752Z
Publisher: Tech & Business (techandbusiness.org)
