benchmarks.site

Public leaderboard · Agents, tool use & computer use

OSWorld (OSWorld-Verified)

Gives an AI control of a real computer (mouse, keyboard, screen) and asks it to do everyday jobs in apps like a browser, spreadsheet or image editor. It shows how close AI is to operating a PC the way a person does.

Leaderboard

Top 3 right now

  1. 1stIntelligence-Indeed Agent90.2
  2. 2ndclaude-fable-5[1m]86.0
  3. 3rdPointer Agent w/ Opus 4.783.6

What it measures

369 tasks in real desktop applications on a virtual machine, checked by execution-based scripts, scored by success rate.

How to read the numbers

Share of the 369 desktop tasks the agent completed successfully (%), averaged over runs where several were submitted — higher is better.

Results

25 rows · higher is better (Success rate (%))CSV
1Intelligence-Indeed AgentIntelligence IndeedAgentic framework10090.22026-07-25
2claude-fable-5[1m]AnthropicGeneral model10086.02026-08-01
3Pointer Agent w/ Opus 4.7PointerAgentic framework10083.62026-05-21
4claude-opus-5[1m]AnthropicGeneral model10083.42026-08-01
5Coasty CUA v1Coasty TeamGeneral model10082.82026-07-01
6Pointer Agent w/ Sonnet 4.6PointerAgentic framework10081.52026-05-21
7Muse Spark 1.1Meta Superintelligence LabsGeneral model10080.72026-07-09
8Holo3-35B-A3BH CompanySpecialized model10080.4±2.22026-04-20
9OpenAPA w/ gemini-3.1-proLaiyeAgentic framework10078.32026-04-17
10VLAA-GUI w/ Opus 4.5UCSC VLAA LabAgentic framework10076.32026-04-20
11MiniMax M3MiniMaxGeneral model10075.22026-06-08
12HIPPO Agent w/ Opus 4.5LenovoAgentic framework10074.52026-02-25
13Qwen 3.7 PlusQwen Team, Alibaba GroupGeneral model10073.32026-05-25
14Kimi K2.6Moonshot AIGeneral model10073.12026-04-20
15agent s3 w/ Opus 4.5 + GPT-5 bBoN (N=10)SimularAgentic framework10072.62025-12-11
16claude-sonnet-4-6AnthropicGeneral model10072.12026-03-08
17agent s3 w/ GPT-5 bBoN (N=10)SimularAgentic framework10069.92025-10-04
18UiPath Screen Agent w/ Opus 4.5UiPathAgentic framework10067.12025-12-24
19agent s3 w/ Opus 4.5 bBoN (N=1)SimularAgentic framework10066.0±1.12025-12-11
20OS-Symphony w/ GPT-5Shanghai AI LaboratoryAgentic framework5065.82026-01-04
21UiPath Screen Agent w/ Opus 4.5UiPathAgentic framework5064.42025-12-24
22GBOX AgentGBOX.AIAgentic framework1564.22025-11-25
23GTA1 w/ GPT-5Salesforce & The Australian National University & The University of Hong KongAgentic framework10063.42025-10-03
24Kimi K2.5Moonshot AIGeneral model10063.32026-01-31
25claude-sonnet-4-5-20250929AnthropicGeneral model10062.92025-10-31

Note: OSWorld-Verified results file behind the official leaderboard. Rows are grouped by model + max-steps setting and averaged the same way the official page does (mean, with ± population std when several runs exist), sorted by score; all approach types (general models, specialized models, agentic frameworks) included. as_of = latest result date in the file.

Source: OSWorld team (XLANG Lab, University of Hong Kong) · os-world.github.io · results as of 2026-08-01 · retrieved 2026-11-17

Markdown · More Agents, tool use & computer use