CloudOpsBench.

CloudOpsBench

A benchmark for AI agents on real cloud operations

How it worksView the tasks
Model leaderboard. Select a row to view details.
RankModelHarnessResolution rate Release date Time Tokens Cost
01claude-fable-5-1 (max)claude-codeclaude-code
60.6%
38.6h97.9M$404.3
02gpt-6-astra (max)codexcodex
58.9%
16.5h98.0M$310.9
03gemini-4-argon (high)gemini-cligemini-cli
57.5%
34.2h173.9M$100.5
04grok-4.7 (high)cursor-clicursor-cli
56.5%
33.2h178.3M$138.8
05gpt-6-astra (high)codexcodex
54.2%
16.1h62.4M$188.4
06grok-4.7 (high)grok-buildgrok-build
48.3%
25.9h173.4M$127.3
07claude-opus-5-5 (high)terminus-2terminus-2
47.8%
19.5h22.6M$45.1
08claude-opus-5-5 (high)claude-codeclaude-code
46.7%
14.0h37.4M$36.6
09claude-fable-5-1 (high)terminus-2terminus-2
43.8%
19.0h37.3M$319.1
10gemini-3.8-flash (high)terminus-2terminus-2
42.8%
33.2h337.4M$70.0
11gpt-6-astra (high)terminus-2terminus-2
42.8%
12.9h19.2M$94.6
12grok-4.7 (medium)grok-buildgrok-build
40.0%
21.4h121.6M$91.6
13grok-4.6 (high)cursor-clicursor-cli
39.7%
19.7h78.2M$63.8
14grok-4.6 (medium)grok-buildgrok-build
36.7%
16.2h59.4M$51.2
15grok-4.6 (high)grok-buildgrok-build
36.1%
17.7h72.0M$61.7
16qwen3.8-max (high)terminus-2terminus-2
33.7%
79.1h78.4M$61.8
17deepseek-v4.1-flash (high)terminus-2terminus-2
21.8%
22.2h68.1M$7.5
18kimi-k3.1-pro (high)terminus-2terminus-2
19.1%
14.8h7.4M$10.7