VibePrevisBench - Screen - Photo: leaderboard
VibePrevisBench - Screen - Photo. VibePrevisBench frozen paper Table1:242 Blender4D tasks,94script39photo23continuation86video,agent writes/runs Blender Python. Scene score is rule-based84,406hidden facts; Screen score Gemini4Argon VLM judge (photo recognition only). Failed runs score0. All is weighted native aggregate; overall mean of SceneAll andScreenAll. Harness and exact effort are observation settings, not checkpoints. README values are native one-decimal report; JSON provides two-decimal aggregate separately. Metric: Score (%). Higher is better.
Metric: Score (%). Source: raw.githubusercontent.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6.1 Sol | 93.8 | #10 |
| 2 | GPT-6 Astra | 93.6 | #6 |
| 3 | GPT-5.6 Sol | 92.5 | #30 |
| 4 | Claude Opus 5.5 | 90.4 | #5 |
| 5 | Gemini 4 Argon | 90.3 | #11 |
| 6 | DeepSeek V4.1 Flash | 90.2 | #59 |
| 7 | Gemini 3.8 Flash | 87.7 | #36 |
| 8 | Seed 2.1 Pro | 86.8 | #135 |
| 9 | GLM-5.3 Flash | 84.8 | #122 |
| 10 | Muse Spark 1.3 | 84.7 | #46 |
| 11 | Kimi K3 | 81.2 | #54 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=vibeprevisbench-screen-photo · How It Works · Data refreshed daily, snapshot 2026-11-13.