5 published experiments, 189 recorded runs across games, graphics, CAD and robot simulation. Newest first.
There was no universal winner. Flash-model performance was brutally task-specific.
Games · Graphics · CAD/3D · Robot simulation14 tests · 54 recorded runsGeneral Library · no campaign package
Opus 5 delivered the strongest detail in several rounds. It also made the efficiency argument against itself.
Games · Interactive worlds9 tests · 32 recorded runsFull reproduction path
Kimi K3 was more consistent; Qwen's visual-agent workflow was not reliable enough to win overall.
Games · Interactive worlds9 tests · 42 recorded runsNot released
Fable 5 sets the visual ceiling. Kimi K3 wins on value. GPT-5.6 Sol Ultra is the practical middle when technical coherence matters without paying Fable’s full cost.
Games · Interactive worlds9 tests · 33 recorded runsProject + prompt available
Fable sets the visual ceiling. Sol Ultra wins the value argument.
Games · Interactive worlds7 tests · 28 recorded runsPrompt only
Each listed result is one recorded attempt. Reliability is not established. Costs retain the estimate type and caveats recorded in the corresponding ledger.
