Best LLM for 3D scenes
- GPT-6 Astra leads
- 1500
- models ranked
- 116
- lead over #2
- +24 Elo
Design Arena: 3D · read October 22
As of October 22, 2026, OpenAI GPT-6 Astra leads 116 models on Design Arena: 3D with 1500, ahead of Anthropic Claude Opus 5.5 (1476) and OpenAI GPT-6.1 Sol (1460).
The full best LLM for web design ranking combines this benchmark with others.
| Rank | Model | Design Arena: 3D | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra OpenAI | 1500 | 85.1 | $50.00 | |
| 2 | Claude Opus 5.5 Anthropic | 1476 | 88.8 | $20.00 | |
| 3 | GPT-6.1 Sol OpenAI | 1460 | 82.4 | $10.00 | |
| 4 | Claude Sonnet 5.5 Anthropic | 1445 | 87.3 | $10.00 | |
| 5 | Muse Wonder Meta | 1427 | 71.9 | $20.00 | |
| 6 | Claude Fable 5.1 Anthropic | 1392 | 84.2 | $50.00 | |
| 7 | Gemini 4 Argon Google | 1388 | 81.0 | $10.00 | |
| 8 | Kimi K3 Moonshot AI | 1387 | 62.3 | $15.00 | |
| 9 | GPT-5.6 Sol OpenAI | 1384 | 73.5 | $20.00 | |
| 10 | DeepSeek V4.1 Flash 1017 DeepSeek | 1358 | 70.3 | $0.60 | |
| 11 | Muse Spark 1.3 Meta | 1358 | 57.4 | $4.25 | |
| 12 | DeepSeek V4.1 Flash DeepSeek | 1355 | 62.5 | $0.84 | |
| 13 | Claude Haiku 5.5 Anthropic | 1342 | 70.7 | $0.50 | |
| 14 | Qwen3.8 Max (0902) Qwen | 1336 | 62.4 | $6.00 | |
| 15 | GLM 5.1 Z.ai | 1336 | 49.2 | $4.40 | |
| 16 | Claude Opus 5 Anthropic | 1334 | 83.0 | $25.00 | |
| 17 | GLM 5.3 Z.ai | 1332 | 70.6 | $4.40 | |
| 18 | MiMo-V2.6-Pro Xiaomi | 1330 | 65.4 | $0.87 | |
| 19 | Claude Fable 5 Anthropic | 1318 | 79.3 | $50.00 | |
| 20 | GLM 5.3 Flash Z.ai | 1316 | 57.7 | $0.50 |
How this list is built
Generated from the dataset on every refresh. Nobody picks the order by hand.
Models by their Design Arena: 3D result, as the board publishes it, best first. Design Arena shows people two designs made by two models from the same prompt (a 3D scene) and asks which is better. The number is the Elo rating built from those votes, as Design Arena publishes it.
- OpenRouter, fetched
- Design Arena, fetched