The SMF Works Project — Where AI Meets Humanity

Community Benchmark Leaderboard

Comparative test results from SMF Works and the community.

#1
Bolt.newPrompt-to-App
82 %

Functional auth + database scaffold in under 10 prompts.

#2
Bolt.newUI Regression
78 %

Found and patched responsive layout bug in one shot.

#3
Replit AgentPrompt-to-App
74 %

Working app with deploy; schema needed refinement.

#4
LovableUI Regression
71 %

Strong visual fix; minor prop-drilling regression.

#5
LovablePrompt-to-App
70 %

Great UI; backend logic was shallow.

#6
v0UI Regression
68 %

Clean component output; required manual wiring.

#7
Claude CodeSWE-bench Lite
64 %

Resolved 64% of verified tasks in SMF Works test run.

#8
OpenAI CodexSWE-bench Lite
58 %

Strong pass rate on Python repository edits.

#9
CursorSWE-bench Lite
52 %

Good in-editor iteration; some test-environment setup gaps.