OS-Kairos
Agents predict Android actions for tasks such as finding a product in Amazon. The task succeeds when every action matches the recorded trajectory.
The DKR-GUI team reported separate action and grounding scores on Kairos. The original test split has 200 trajectories covering 198 goals.
- Made by
- Shanghai Jiao Tong University and Meta
- Released
- February 2025
- Kind
- Phones
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- Last model added
- August 2024
- Scores refreshed
- November 20th, 2026
OS-Kairos tops the leaderboard with 88.2% Qwen-VL-Max leads with 29.8%
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen-VL-Max, at 29.8%. This view shows only models we track, each at its best.
- 1 OS-Kairos Zero-shot 88.2%
- 2 GPT-4o (snapshot unspecified) Zero-shot 39.1%
- 3 Qwen-VL-Max Zero-shot 29.8%
- 4 OS-Atlas-Pro-7B Zero-shot 14.3%
- 5 GLM-4V-Plus Zero-shot 4.35%
- 6 Auto-UI Zero-shot 0%
- 7 Qwen2-VL-7B-Instruct Zero-shot 0%
Source: The OS-Kairos paper
- 1 Qwen-VL-Max Zero-shot 29.8%
- 2 Qwen2-VL-7B-Instruct Zero-shot 0%
Source: The OS-Kairos paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | OS-Kairos | 88.2% | Zero-shot | Official leaderboard | |
| 2 | GPT-4o (snapshot unspecified) | 39.1% | Zero-shot | Official leaderboard | |
| 3 | Qwen-VL-Max | 29.8% | Zero-shot | Jan 2024 | Official leaderboard |
| 4 | OS-Atlas-Pro-7B | 14.3% | Zero-shot | Official leaderboard | |
| 5 | GLM-4V-Plus | 4.35% | Zero-shot | Official leaderboard | |
| 6 | Auto-UI | 0% | Zero-shot | Official leaderboard | |
| 7 | Qwen2-VL-7B-Instruct | 0% | Zero-shot | Aug 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen-VL-Max | 29.8% | Zero-shot | Jan 2024 | Official leaderboard |
| 2 | Qwen2-VL-7B-Instruct | 0% | Zero-shot | Aug 2024 | Official leaderboard |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official OS-Kairos leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Computer and browser use
- BrowseComp In use Hard-to-find facts that need persistent web browsing. Short verifiable answers graded for exact correctness. Claude Fable 5 93.3%
- ScreenSpot-Pro In use GUI grounding on high-resolution professional-software screenshots. The model must locate the target UI element. GPT-6 Astra 92.7%
- AndroidWorld In use Agents complete 116 tasks across 20 real Android apps in an emulator. Each task checks the device's system state to judge success. GPT-6 Astra 100%
- OSWorld 2.1 In use Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. Claude Opus 5 44.3%
- WebArena-Verified In use Agents complete 812 audited WebArena tasks on self-hosted websites. Deterministic evaluators check the agent's answer and captured network traces.
Muse Spark 1.1 69%
- OSWorld 2.1 (offline subset) In use Agents complete the 82 OSWorld 2.1 tasks on the official offline list, with no internet access. Scored by checkpoint partial credit and strict pass rate. Claude Fable 5.2 89.3%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in August 2024, for Qwen2-VL-7B-Instruct. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (7 entries).
- Pass Models scored 2 models we track, plus 5 other entries on its leaderboard.
- Warning Newest score The newest came 2 years ago, in August 2024, for Qwen2-VL-7B-Instruct.
- Note Frontier models None of its models are on the capabilities index.
- Pass Headroom The best is 29.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 2 rows for models we track and 5 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.