Benchmarks / 3D
AA-3D
AA-3D: agents in an offline sandbox turn a written brief into an interactive three.js scene; Elo from blind pairwise votes by AA's recruited panel.
AA-3D
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Sonnet 5.5 | max | 1197 |
| 2 | Opus 5.5 | max | 1191 |
| 3 | GPT-6 Astra | max | 1166 |
| 4 | Gemini 4 | high | 1154 |
| 5 | Fable 5.1 | max | 1133 |
| 6 | GPT-6 Sol | max | 1123 |
| 7 | GPT-6.1 Sol | max | 1118 |
| 8 | Muse Spark 1.3 | max | 1037 |
| 9 | GPT-6 Luna | max | 1000 |
| 10 | Haiku 5.5 | max | 997 |
| 11 | Grok 4.7 | xhigh | 995 |
| 12 | Kimi K3 | max | 981 |
| 13 | MiMo V2.6 Pro | – | 980 |
| 14 | Gemini 3.8 Flash | high | 969 |
| 15 | Qwen 3.8 Max | max | 948 |
| 16 | GLM 5.3 Flash | max | 935 |
| 17 | Mistral Large 4 | – | 870 |
| 18 | Step 5 Preview | – | 852 |
| 19 | GLM 5.3 | max | 787 |
| 20 | MiniMax M3 | – | 740 |
| 21 | Inkling | xhigh | 581 |
| 22 | Nemotron 3 Ultra | max | 569 |