VoxelBench (text)

Models build 3D voxel scenes from text prompts; humans vote head-to-head on the renders. Glicko-2 rating.

54 models · 54 results · data as of 2026-11-03 · voxelbench.ai ↗

All results

54 rows

#	Model	Effort	VoxelBench (text)
1	GPT-6 Astra	max	2690 ±265
2	GPT-6.1 Sol	max	2625 ±220
3	Opus 5.5	max	2620 ±127
4	GPT-6 Sol	max	2413 ±122
5	Opus 5	max	2294 ±110
6	GPT-5.6 Sol	max	2288 ±106
7	GPT-6.1 Terra	max	2276 ±88
8	Gemini 4 Argon	max	2234 ±108
9	Kimi K3.1	max	2224 ±76
10	Fable 5	max	2116 ±114
11	DeepSeek V4.1 Flash 1017	max	2038 ±104
12	Grok 4.6	xhigh	2029 ±108
13	GPT-5.5	xhigh	1946 ±104
14	GPT-5.5 Pro	xhigh	1931 ±112
15	Kimi K3	max	1922 ±104
16	GPT-5.6 Terra	max	1893 ±104
17	DeepSeek V4.1 Flash	max	1870 ±112
18	GPT-5.6 Luna	max	1856 ±106
19	Gemini 3.7 Flash	max	1844 ±104
20	GPT-6 Luna	max	1806 ±112
21	Gemini 3.8 Flash	max	1803 ±112
22	MiniMax M3.1	max	1798 ±106
23	GLM 5.3	max	1791 ±106
24	Qwen 3.8 Max	max	1778 ±104
25	Gemini 3.6 Flash	xhigh	1762 ±106
26	Grok 4.5	max	1729 ±104
27	Haiku 5.5	max	1725 ±110
28	GLM 5.3 Flash	max	1725 ±104
29	Gemini 3.1 Pro	high	1643 ±110
30	Gemini 3.5 Flash	xhigh	1636 ±104