Tools & AgentsTianbao Xie et al. (University of Hong Kong, Salesforce Research and others)

OSWorld published

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

OSWorld published — a multimodal agent benchmark with execution-based evaluation in real OS environments

Why it matters

The de facto standard benchmark for computer-use agents

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2404.07972

Added in Week 35, 2026 · See that week

On the map — ③ Tools & Agents line

  1. Previous stopGAIA benchmark
  2. This recordOSWorld
  3. Next stopClaude tool use
See it on the map

Same week

Controversy spreads over whether Devin's Upwork demo holds upPressxAI previews Grok-1.5VOfficialCohere launches Rerank 3OfficialRULER publishedPaper

Nearby on the same axis

Controversy spreads over whether Devin's Upwork demo holds upPressOpenAI ends the ChatGPT plugins beta, replacing it with GPTsOfficialGitHub launches Copilot Workspace technical previewOfficialSWE-agentPaper