Search-to-World - World Delivery: leaderboard
World Delivery Rate on 300 requests: percentage delivering a 3D world that passes request-alignment, spatial-extent and perceptual-quality gates. Each evaluated model fills all four primary agent roles.
Metric: World Delivery Rate (%). Source: arxiv.org. Saturation forecast: Not forecast. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Flash | 78.7 | #155 |
| 2 | GPT-5.6 Sol | 53 | #22 |
| 3 | Kimi K3 | 52.7 | #35 |
| 4 | GLM-5.2 | 52.7 | #79 |
| 5 | GPT-5.6 Terra | 51 | #41 |
| 6 | DeepSeek V4 Pro | 48.3 | #112 |
| 7 | Qwen 3.5 397B A17B | 44 | #154 |
| 8 | Qwen 3.5 35B A3B | 39.7 | #268 |
| 9 | Qwen 3.5 4B | 38 | #515 |
| 10 | Qwen 3.5 9B | 29.3 | #391 |
| 11 | Claude Fable 5 | 29.3 | #20 |
| 12 | Claude Opus 5 | 16 | #16 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=search-to-world-world-delivery · How It Works · Data refreshed daily, snapshot 2026-11-06.