Skip to content
gentic.news — AI News Intelligence Platform
Connecting to the Living Graph…
Screen-level OS controlOpen Source#9 of 14 in category

GUI-Owl-1.5

Alibaba · Launched Feb 2026

Alibaba's Mobile-Agent-v3.5 GUI model; OSWorld-Verified 52.9% at 8B scale, AndroidWorld 71.6%, ScreenSpot-Pro 80.3% — SOTA among open GUI models.

3
Benchmarks scored
80.3
Peak score
0
Article mentions
Yes
Open source

Benchmark performance

ScreenSpot-Pro

High-resolution professional-software GUI grounding (CAD, IDEs, creative suites) - far harder than consumer ScreenSpot.

80.3
★ SOTA
AndroidWorld

Google's 116 hand-built tasks across 20 real Android apps on a live emulator. Mobile is harder than desktop for most agents.

71.6
Gap to SOTA: -15.5pp (held by Surfer 2)Human: 80%Benchmark docs →
OSWorld-Verified

369 real tasks on a live Ubuntu desktop VM: file I/O, spreadsheets, creative apps, settings. The July 2025 Verified rebuild moved to AWS (50x parallel) and fixed 300+ task bugs. The flagship computer-use benchmark.

52.9
Gap to SOTA: -30.5pp (held by Claude Opus 4.8)Human: 72.4%Benchmark docs →

Other screen-level os control agents

The 14 agents in this category, ranked by peak benchmark.

AgentMakerLaunchPeakPricing
Claude Sonnet 4.6Anthropic2026-021470.0$3 / $15 per M tokens
Claude Opus 4.8Anthropic2026-0588.6$5 / $25 per M tokens
UI-TARS-2OSSByteDance2025-0988.2Open weights
Claude Opus 4.7Anthropic2026-0387.6$5 / $25 per M tokens
Claude Mythos PreviewAnthropic2026-0486.9Research preview
Gemini 3.1 ProGoogle DeepMind2026-0285.9Google API
GPT-5.5OpenAI2026-0582.6OpenAI API
Holo3-35B-A3BH Company2026-0482.6H Company
Holo3-122B-A10BH Company2026-0478.8H Company
GPT-5.4OpenAI2026-0375.0OpenAI API

Quick facts

Type
Screen-level OS control
Maker
Alibaba
Launch
2026-02-01
Open source
Yes
Pricing
Open weights
Benchmarks scored
3
Article mentions
0
Rank in category
#9 of 14