Visual benchmark ranking
168 working runs across 16 shared briefs. Open a model to inspect the output.
S
Start here for serious builds.
Best hard-build performer when access is available.
9 published runs →
Strong technical all-rounder for demanding interactive builds.
10 published runs →
Complete playable scenes across all 16 prompts, with committed visual direction.
16 published runs →A
Strong choices that are very economically efficient.
Excellent end-to-end builds with strong visual discipline.
10 published runs →All 16 prompts delivered as working, validated builds in a single session.
16 published runs →Reliable, polished default for complex product work.
11 published runs →Polished interactive games across its first six published runs.
6 published runs →
Extremely cheap API costs and decent performance.
16 published runs →B
Useful specialists that need review.
Fast, capable workhorse with strong overall execution.
9 published runs →Good rapid builds, although quality is less consistent.
7 published runs →
Strong with a clear plan and a capable coding harness.
11 published runs →Very fast and useful, but complex builds need closer review.
9 published runs →Excellent planner that can overthink and drift on long builds.
9 published runs →Coherent, functional scenes with less finish than the leaders.
9 published runs →Useful low-cost scaffolding that needs stronger-model review.
11 published runs →C
Inconsistent in the current run set.