# threejseval — AI model rankings

Every model gets the same prompt — "the Eiffel Tower", "a glass fishbowl", "a robot arm picking toys into a box" — and builds a 3D scene in Three.js. Real people then see two scenes side by side, names hidden, and vote for the one they prefer. The votes become a chess-style Elo rating on threejseval.com, averaged across all the prompts. It measures whether the scene looks and moves right to a human eye, not whether the code passes a test. Higher is better.

Scores from [threejseval](https://threejseval.com/ranking), which runs the benchmark and publishes the full field.

29 tracked models have a published threejseval score. Higher is better. Scores are gathered from threejseval; recorded retrieval dates appear beside the scores.

## Ranking

| Rank | Model | Developer | Score | Source | Released |
| --- | --- | --- | --- | --- | --- |
| 1 | Claude Opus 5.5 | Anthropic | 2065 | [threejseval](https://threejseval.com/ranking) | Sep 22 2026 |
| 2 | GPT-6.1 Sol | OpenAI | 1972 | [threejseval](https://threejseval.com/ranking) | Sep 29 2026 |
| 3 | GPT-6 Astra | OpenAI | 1876 | [threejseval](https://threejseval.com/ranking) | Sep 3 2026 |
| 4 | Claude Fable 5.1 | Anthropic | 1858 | [threejseval](https://threejseval.com/ranking) | Sep 1 2026 |
| 5 | Claude Sonnet 5.5 | Anthropic | 1836 | [threejseval](https://threejseval.com/ranking) | Sep 28 2026 |
| 6 | Gemini 4 Argon | Google | 1755 | [threejseval](https://threejseval.com/ranking) | Sep 30 2026 |
| 7 | DeepSeek-V4.1-Flash-1017 | DeepSeek | 1749 | [threejseval](https://threejseval.com/ranking) | Oct 17 2026 |
| 8 | Claude Opus 5 | Anthropic | 1707 | [threejseval](https://threejseval.com/ranking) | Jul 24 2026 |
| 9 | DeepSeek-V4.1-Flash | DeepSeek | 1642 | [threejseval](https://threejseval.com/ranking) | Sep 10 2026 |
| 10 | GPT-6 Sol | OpenAI | 1622 | [threejseval](https://threejseval.com/ranking) | Sep 22 2026 |
| 11 | Grok 4.7 | SpaceXSI | 1618 | [threejseval](https://threejseval.com/ranking) | Sep 21 2026 |
| 12 | Gemini 3.8 Flash | Google | 1564 | [threejseval](https://threejseval.com/ranking) | Sep 2 2026 |
| 13 | Kimi K3 | Moonshot AI | 1512 | [threejseval](https://threejseval.com/ranking) | Jul 16 2026 |
| 14 | Claude Fable 5 | Anthropic | 1511 | [threejseval](https://threejseval.com/ranking) | Jun 9 2026 |
| 15 | Gemini 3.7 Flash | Google | 1468 | [threejseval](https://threejseval.com/ranking) | Aug 13 2026 |
| 16 | Grok 4.6 | SpaceXSI | 1453 | [threejseval](https://threejseval.com/ranking) | Aug 12 2026 |
| 17 | Muse Spark 1.3 | Meta | 1432 | [threejseval](https://threejseval.com/ranking) | Sep 2 2026 |
| 18 | GPT-5.6 Terra | OpenAI | 1429 | [threejseval](https://threejseval.com/ranking) | Jun 26 2026 |
| 19 | Claude Haiku 5.5 | Anthropic | 1401 | [threejseval](https://threejseval.com/ranking) | Oct 7 2026 |
| 20 | GLM-5.3-Flash | Z.ai | 1373 | [threejseval](https://threejseval.com/ranking) | Aug 26 2026 |
| 21 | GPT-5.6 Sol | OpenAI | 1369 | [threejseval](https://threejseval.com/ranking) | Jun 26 2026 |
| 22 | GPT-6 Luna | OpenAI | 1365 | [threejseval](https://threejseval.com/ranking) | Sep 22 2026 |
| 23 | Gemini 3.6 Flash | Google | 1340 | [threejseval](https://threejseval.com/ranking) | Jul 21 2026 |
| 24 | GPT-5.6 Luna | OpenAI | 1270 | [threejseval](https://threejseval.com/ranking) | Jun 26 2026 |
| 25 | Claude Sonnet 5 | Anthropic | 1204 | [threejseval](https://threejseval.com/ranking) | Jun 30 2026 |
| 26 | GPT-5.3-Codex | OpenAI | 1148 | [threejseval](https://threejseval.com/ranking) | Feb 5 2026 |
| 27 | Claude Opus 4.5 | Anthropic | 1138 | [threejseval](https://threejseval.com/ranking) | Nov 24 2025 |
| 28 | Composer 2.5 | SpaceXSI | 1077 | [threejseval](https://threejseval.com/ranking) | May 18 2026 |
| 29 | Mistral Large 4 | Mistral | 1037 | [threejseval](https://threejseval.com/ranking) | Oct 6 2026 |


---

Canonical page: https://aireleasetracker.com/benchmark/threejseval
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.