AI / FlutterBench / Tasks / Build Command-Line CLI App
Dart Utilities

Build Command-Line CLI App

ID: google/dart-build-cli-app

Create a robust Dart command-line interface application with argument parsing, formatted output, and exit code handling.

Cross-Model Benchmark Trials

Comparison of agent configurations evaluated on this task:

Model & HarnessToolingStatusComposite RewardActions
Claude 3.7 Sonnetclaude-code
Dart MCP + Skills PASS 0.97 Inspect Trial Details →
GPT 5codex-agent
Dart MCP + Skills PASS 0.98 Inspect Trial Details →
o3codex-agent
Dart MCP + Skills PASS 0.97 Inspect Trial Details →
Claude 3.5 Haikuclaude-code
Baseline PARTIAL 0.56 Inspect Trial Details →
Gemini 3.1 Flash Litegemini-cli
Baseline PARTIAL 0.39 Inspect Trial Details →
Claude 3.5 Sonnetclaude-code
Baseline PARTIAL 0.79 Inspect Trial Details →
GPT 4ocodex-agent
Baseline PARTIAL 0.71 Inspect Trial Details →
DeepSeek V3deepseek-cli
Baseline PARTIAL 0.71 Inspect Trial Details →
Gemini 3.5 Proantigravity-sdk
Dart MCP + Skills PASS 0.98 Inspect Trial Details →
DeepSeek R1deepseek-cli
Dart MCP + Skills PASS 0.99 Inspect Trial Details →
Gemini 3.5 Flashantigravity-sdk
Dart MCP + Skills PASS 0.93 Inspect Trial Details →
GPT 4o Minicodex-agent
Baseline PARTIAL 0.48 Inspect Trial Details →
DeepSeek Coder V2deepseek-cli
Baseline PARTIAL 0.61 Inspect Trial Details →