Fara-1.5: Scalable Learning Environments for Computer Use Agents

2026-06-18

AI Summary

Manually collecting demonstration data for computer operation tasks is costly and difficult to scale. To address this challenge, this work proposes FaraGen1.5—a modular and scalable data generation framework that integrates real and synthetic environments, multi-model solvers (including GPT-5.4), user simulators, and a triple verification mechanism assessing task correctness, efficiency, and adherence to critical steps. The framework supports simulation of scenarios involving authentication constraints and irreversible operations and iteratively optimizes the composition of training data to enhance quality. Models trained with this framework—Fara1.5 (4B/9B/27B)—achieve state-of-the-art performance on web interaction benchmarks: the 9B variant scores 63.4% on Online-Mind2Web and 86.6% on WebVoyager, while the 27B model reaches 72.3%, rivaling significantly larger closed-source systems.

Abstract

Collecting computer use data from human demonstrations is expensive and slow, motivating the need for scalable generation strategies. This requires two key ingredients: environments in which agents can act and verifiers that can judge whether their demonstrations succeeded. We introduce FaraGen1.5, a scalable data pipeline for computer use agents composed of three modular components: environments, solvers, and verifiers. FaraGen1.5 uses both live websites and synthetic environments that faithfully simulate domains gated by authentication or that require irreversible actions. It employs a solver harness that can be powered by multiple models, including strong frontier models such as GPT-5.4, and also incorporates a user simulator to enable multi-turn rollouts. Finally, FaraGen1.5 scores the resulting trajectories with three complementary verifiers covering task correctness, efficiency, and critical-point adherence. Using data produced by this pipeline, we train Fara1.5, a family of native computer use agents (CUAs) at three scales built on Qwen3.5 (4B, 9B, and 27B). To train these models, we employ a supervised finetuning (SFT) recipe that carefully balances data from FaraGen1.5 for broad coverage, specific high-value tasks, and target model deficiencies in an iterative approach. Each model sets a new state of the art for its size class on browser-use benchmarks: Fara1.5-9B reaches 63.4% on Online-Mind2Web and 86.6% on WebVoyager, while Fara1.5-27B achieves 72.3% on Online-Mind2Web, which is competitive with much larger proprietary systems.

Problem

• computer use agents
• scalable data generation
• human demonstrations
• learning environments
• data collection

Innovation

• scalable data generation
• computer use agents
• synthetic environments
• multi-turn rollouts
• modular verification