Benchmark profile
App-Bench
A six-task full-stack web-app benchmark that measures how much required functionality an AI builder or coding assistant delivers from one prompt without human code edits.
How we show App-Bench
We mirror AfterQuery's App-Bench table captured on July 18, 2026 snapshot. It compares 10 app-building systems on 6 full-stack projects using binary functional rubrics and reports the best result from three one-shot attempts per task.
The rows mix hosted app builders with CLI and IDE coding assistants. We preserve those system names and underlying model labels, but keep the table display-only because a tool-level result is not a controlled base-model comparison.
Published percentile score on App-Bench — July 18, 2026 snapshot
BenchLM mirrors the published published percentile score view for App-Bench. Orchids leads the public snapshot at 76.80% , followed by Claude Code (Opus 4.5) (67.50%) and v0 (64.90%). BenchLM does not use these results to rank models overall.
Orchids
Orchids
afterquery/app-bench/orchids
Claude Code (Opus 4.5)
Anthropic
afterquery/app-bench/claude-code-opus-4-5
v0
Vercel
afterquery/app-bench/v0
Published percentile score table (10 models)
ScoreThe published App-Bench snapshot places Orchids first at 76.80%. The third row is 11.90 points behind. The broader top-10 range is 76.80 points, so the table still separates the published systems.
10 models have been evaluated on App-Bench. The benchmark falls in the Coding category. This category carries a 20% weight in BenchLM.ai's overall scoring system. App-Bench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About App-Bench
Year
2025
Tasks
6 full-stack app-building tasks
Format
Best-of-three one-shot feature completion
Difficulty
Production-style full-stack application generation
App-Bench compares five hosted app builders and five CLI or IDE coding assistants. Each system receives three one-shot attempts per task, the best attempt is graded against binary functional requirements, and two developers reconcile disputed grades. We keep these tool-level rows display-only rather than treating them as base-model scores.
BenchLM freshness & provenance
Version
App-Bench 2025
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
FAQ
What does App-Bench measure?
A six-task full-stack web-app benchmark that measures how much required functionality an AI builder or coding assistant delivers from one prompt without human code edits.
Which model leads the published App-Bench snapshot?
Orchids currently leads the published App-Bench snapshot with 76.80% published percentile score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on App-Bench?
10 AI models are included in BenchLM's mirrored App-Bench snapshot, based on the public leaderboard captured on July 18, 2026 snapshot.
The AI models change fast. We track them for you.
A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.