Holo2: Foundational Models for Navigation and Computer Use Agents

Benchmark Model Rank Results
computer-use-agents-on-androidworldHolo2-30B-A3B–Success Rate: 71.6
computer-use-agents-on-osw-g-osworld-gHolo2-235B-A22B–Accuracy: 79.0
computer-use-agents-on-screenspot-v2Holo2-235B-A22B–Accuracy: 95.9
computer-use-agents-on-webarenaHolo2-30B-A3B–Success Rate: 46.3
computer-use-agents-on-webarenaHolo2-4B–Success Rate: 41.0
computer-use-agents-on-webarenaHolo2-8B–Success Rate: 42.2
computer-use-agents-on-webvoyagerHolo2-30B-A3B–Success Rate: 83.0