Atlas
Benchmarks
326 sources · 1,256 models × 2,956 benchmarks · 154,484 scoresCua Driver Bench v2 (Linux, background)
Cua's Cua Driver Bench v2 on Ubuntu 24.04 with XFCE and the xfwm4 window manager, while a simulated user keeps typing in a window in front: eight of its ten desktop tasks, done only through a computer-use driver's tools. A trial passes when the task's checker confirms the apps' final state, no route out of the GUI was used and none of eight disturbances reached the user; xfwm4 gives every new modal dialog the keyboard, so three interruption tasks can't pass here with any driver. Each task runs five seeded trials and the rate is passes over trials.
Models
6
Top score
12.5
Leader
Claude Haiku 5.5
Updated
—
Resolution
—
Saturation
—
No signal or difficulty estimate. Held out of the fit by design.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Claude Haiku 5.5 | — · Cua Driver Bench on AgentEnv; Claude Code 2.1.280 with no --effort passed and its built-in tools limited to Skill and Read; trycua's cua-driver MCP tools alone; 45 turns and 360 seconds per trial | 12.5 | |
| 2 | Claude Opus 5.5 | — · Cua Driver Bench on AgentEnv; Claude Code 2.1.280 with no --effort passed and its built-in tools limited to Skill and Read; trycua's cua-driver MCP tools alone; 45 turns and 360 seconds per trial | 12.5 | |
| 3 | Claude Sonnet 5.5 | — · Cua Driver Bench on AgentEnv; Claude Code 2.1.280 with no --effort passed and its built-in tools limited to Skill and Read; Vercel Labs' agent-desktop 1.2.0 MCP tools alone, headed in the foreground and headless in the background; 45 turns and 360 seconds per trial | 12.5 | |
| 4 | GPT-6.1 Sol | medium · Cua Driver Bench on AgentEnv; Codex CLI 0.147.0 with every built-in Codex tool off; trycua's cua-driver MCP tools alone, whose screenshots Codex doesn't pass to the model; stopped after 45 tool calls or 360 seconds per trial | 12.5 | |
| 5 | GPT-6.1 Terra | medium · Cua Driver Bench on AgentEnv; Codex CLI 0.147.0 with every built-in Codex tool off; trycua's cua-driver MCP tools alone, whose screenshots Codex doesn't pass to the model; stopped after 45 tool calls or 360 seconds per trial | 12.5 | |
| 6 | GPT-6 Luna | medium · Cua Driver Bench on AgentEnv; Codex CLI 0.147.0 with every built-in Codex tool off; trycua's cua-driver MCP tools alone, whose screenshots Codex doesn't pass to the model; stopped after 45 tool calls or 360 seconds per trial | 12.5 |
