Meta researchers taught an 8B AI model to match Claude Opus 4.5 — without the frontier price tag
Consider an AI agent tasked with a complex enterprise workflow like migrating massive batches of customer records from a legacy CRM to a cloud database. The agent cannot rely solely on its internal context window for a job spanning hours and depends on the runtime layer, aka the harness . This harness provides execution feedback, like server logs, to help the agent maintain an accurate…
Meta researchers have developed a new framework called EvoHarness-RL that enables an 8B AI model to effectively utilize an execution harness, similar to Claude Opus 4.5, without the higher costs associated with frontier models. The harness provides feedback, state trackers, and control-flow mechanisms to help AI agents manage complex tasks such as migrating customer records between systems.
EvoHarness-RL introduces a unified Belief, Progress, and Experience (BPE) workspace to help the agent dynamically manage its external environment and leverage historical knowledge. By training the AI to construct and utilize this workspace, researchers aim to reduce engineering resources spent on manual logic and rigid memory structures.
Written by urgent.news from VentureBeat's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
