Perception
Test OCR, small target grounding, coordinate accuracy and response to changing interface layouts.
AGENT USE CASE · SOURCE-FIRST
Computer Use agents.
Computer-use capability requires a model or system to interpret screen state, plan actions and control an interface. Image-input support is not proof that a model can navigate a desktop reliably.
What must be evaluated?
Action chain
Evaluate long-horizon task success on a specific OS/browser setup, with versioned evaluation and explicit failure records.
Safety
Use an isolated desktop, URL allowlists, sensitive-action confirmations, timeouts and full action logs.
Evidence
Keep author benchmark scores, third-party evaluations and independent OMW experiments separate. Unknown must never mean no capability.
8 catalog candidates and components
These model records are included by catalog modality or category, not by a verified task-success benchmark. Follow their primary sources before testing.
MiniMax-M3Core candidate · MiniMax Community LicenseInspect model →MiMo-V2.6-Pro-RLCore candidate · mitInspect model →Kimi-K3Core candidate · otherInspect model →GLM-5.3-FlashCore candidate · mitInspect model →Qwen3.8-27BCore candidate · apache-2.0Inspect model →gemma-4-26B-A4B-itCore candidate · apache-2.0Inspect model →Mistral Small 4 119B A6BCore candidate · apache-2.0Inspect model →Holo4-27BCore candidate · cc-by-nc-4.0Inspect model →
COMMON QUESTIONS
Frequently asked questions
Does a multimodal LLM automatically support computer use?
No. Understanding screenshots and successfully performing GUI actions are distinct, separately testable capabilities.
Can a computer-use agent run offline?
Only if the checkpoint, controller, applications and dependencies work within that offline environment.
READ THE ORIGINALS
Documentation and primary sources
REAL SOURCE EVIDENCE
Publisher-reported task results
Different benchmark versions are not comparable; no numbers below were independently reproduced by Open Model Weights.
Evaluate before deployment.
Model metadata is source-linked; agent task success, tool-call reliability and production readiness require separate tests.
Review the methodology → Return to the explorer →