VoxelBench Text-Prompt Leaderboard (VoxelBench Text)

We show this table for reference; we do not rank on it.

Data verified November 12, 2026

A live human-preference benchmark where language models turn text prompts into voxel structures and voters compare anonymous builds from the same prompt.

Glicko-2 rating on VoxelBench text-prompt leaderboard — November 12, 2026

We mirror the published glicko-2 rating view for VoxelBench text-prompt leaderboard. GPT-6 Astra leads the public snapshot at 2724, followed by Claude Opus 5.5 (2578) and GPT-6.1 Sol (2532). We do not use these results to rank models overall.

1Closed

GPT-6 Astra

OpenAI

5,728 votes · 96.2% win · 95% CI 2461–2987

2724

Overall 85.65Context 1.05M

2Closed

Claude Opus 5.5

Anthropic

2,826 votes · 93.8% win · 95% CI 2451–2705

2578

Overall 86.62Context 1M

3Closed

GPT-6.1 Sol

OpenAI

2,307 votes · 91.6% win · 95% CI 2309–2755

2532

Overall 84.6Context 1.05M

57 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 12, 2026

Glicko-2 rating table (57 models)

Glicko-2 rating results for VoxelBench Text
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	GPT-6 AstraOpenAI5,728 votes · 96.2% win · 95% CI 2461–2987	2724	Not reported	Closed
2	Claude Opus 5.5Anthropic2,826 votes · 93.8% win · 95% CI 2451–2705	2578	Not reported	Closed
3	GPT-6.1 SolOpenAI2,307 votes · 91.6% win · 95% CI 2309–2755	2532	Not reported	Closed
4	GPT-6 SolOpenAI1,105 votes · 82.3% win · 95% CI 2237–2491	2364	Not reported	Closed
5	Claude Opus 5Anthropic3,559 votes · 83.3% win · 95% CI 2189–2429	2309	Not reported	Closed
6	GPT-6.1 TerraOpenAI1,545 votes · 82.9% win · 95% CI 2238–2368	2303	Not reported	Closed
7	GPT-5.6 SolOpenAI5,129 votes · 85.2% win · 95% CI 2142–2358	2250	Not reported	Closed
8	Kimi K3.1 ProMoonshot AI1,650 votes · 79.5% win · 95% CI 2079–2275	2177	Not reported	Open
9	Gemini 4 ArgonGoogle1,970 votes · 80.4% win · 95% CI 2036–2268	2152	Not reported	Closed
10	Claude Fable 5Anthropic5,457 votes · 84.2% win · 95% CI 1994–2218	2106	Not reported	Closed
11	Grok 4.6xAI3,080 votes · 71.2% win · 95% CI 1967–2187	2077	Not reported	Closed
12	DeepSeek-v4.1-Flash-1017DeepSeek1,161 votes · 70.8% win · 95% CI 1926–2142	2034	Not reported	Unknown
13	GPT-5.5 ProOpenAI6,127 votes · 80.8% win · 95% CI 1854–2098	1976	Not reported	Closed
14	GPT-5.6 TerraOpenAI2,959 votes · 60.5% win · 95% CI 1820–2032	1926	Not reported	Closed
15	GPT-5.5OpenAI6,834 votes · 80.0% win · 95% CI 1806–2030	1918	Not reported	Closed
16	Kimi K3Moonshot AI4,377 votes · 70.9% win · 95% CI 1798–2014	1906	Not reported	Open
17	DeepSeek V4.1 Flash 1017DeepSeek2,902 votes · 56.5% win · 95% CI 1758–2002	1880	Not reported	Open
18	Gemini 3.7 FlashGoogle3,200 votes · 57.1% win · 95% CI 1764–1984	1874	Not reported	Closed
19	space-bunny-alphaUnknown991 votes · 48.8% win · 95% CI 1755–1975	1865	Not reported	Unknown
20	Muse WonderMeta1,051 votes · 59.1% win · 95% CI 1716–1940	1828	Not reported	Closed
21	GPT-5.6 LunaOpenAI3,032 votes · 54.7% win · 95% CI 1710–1934	1822	Not reported	Closed
22	GPT-6 LunaOpenAI1,053 votes · 47.6% win · 95% CI 1704–1940	1822	Not reported	Closed
23	MiniMax M3.1 FlashMiniMax983 votes · 54.4% win · 95% CI 1702–1934	1818	Not reported	Closed
24	Gemini 3.8 FlashGoogle2,753 votes · 47.9% win · 95% CI 1684–1920	1802	Not reported	Closed
25	Gemini 3.6 FlashGoogle3,519 votes · 52.8% win · 95% CI 1681–1893	1787	Not reported	Closed
26	GLM-5.3Z.AI3,207 votes · 50.3% win · 95% CI 1669–1889	1779	Not reported	Open
27	Qwen3.8 MaxAlibaba3,201 votes · 55.6% win · 95% CI 1638–1866	1752	Not reported	Open
28	GLM-5.3-Flash (Max)Z.AI3,787 votes · 51.7% win · 95% CI 1635–1855	1745	Not reported	Unknown
29	Grok 4.5xAI4,062 votes · 54.4% win · 95% CI 1631–1851	1741	Not reported	Closed
30	Claude Haiku 5.5Anthropic836 votes · 47.5% win · 95% CI 1603–1847	1725	Not reported	Closed
31	Gemini 3.5 FlashGoogle4,471 votes · 54.0% win · 95% CI 1534–1758	1646	Not reported	Closed
32	Gemini 3.1 Pro PreviewGoogle7,629 votes · 72.0% win · 95% CI 1505–1733	1619	Not reported	Unknown
33	DeepSeek V4 Pro 0813DeepSeek2,859 votes · 39.8% win · 95% CI 1508–1728	1618	Not reported	Open
34	Qwen3.7 MaxAlibaba4,073 votes · 47.4% win · 95% CI 1497–1725	1611	Not reported	Closed
35	Claude Opus 4.8 (Max)Anthropic4,343 votes · 52.1% win · 95% CI 1477–1709	1593	Not reported	Unknown
36	Muse Spark 1.3Meta2,106 votes · 42.3% win · 95% CI 1465–1719	1592	Not reported	Closed
37	GLM-5.2Z.AI4,148 votes · 45.4% win · 95% CI 1465–1705	1585	Not reported	Open
38	Claude Sonnet 5 (xhigh)Anthropic3,373 votes · 41.8% win · 95% CI 1430–1658	1544	Not reported	Unknown
39	Muse Spark 1.1Meta3,316 votes · 34.8% win · 95% CI 1426–1654	1540	Not reported	Closed
40	Gemini 2.5 Deep ThinkGoogle6,956 votes · 70.5% win · 95% CI 1335–1589	1462	Not reported	Unknown
41	DeepSeek V4 Flash 0731DeepSeek3,169 votes · 28.4% win · 95% CI 1335–1563	1449	Not reported	Open
42	Gemini 3 Deep ThinkGoogle4,156 votes · 48.8% win · 95% CI 1316–1578	1447	Not reported	Unknown
43	GLM-5.1Z.AI4,928 votes · 38.6% win · 95% CI 1327–1559	1443	Not reported	Open
44	GPT-5.4OpenAI5,186 votes · 50.8% win · 95% CI 1303–1553	1428	Not reported	Closed
45	Claude Opus 4.6Anthropic4,963 votes · 50.0% win · 95% CI 1290–1548	1419	Not reported	Closed
46	GPT-5.2 (xHigh)OpenAI4,119 votes · 51.8% win · 95% CI 1272–1542	1407	Not reported	Unknown
47	MiniMax M3MiniMax4,198 votes · 34.4% win · 95% CI 1278–1524	1401	Not reported	Open
48	Claude Opus 4.7Anthropic4,382 votes · 39.5% win · 95% CI 1266–1520	1393	Not reported	Closed
49	Claude Opus 4.5Anthropic4,486 votes · 49.8% win · 95% CI 1226–1488	1357	Not reported	Closed
50	Gemini 3 Pro PreviewGoogle4,316 votes · 46.3% win · 95% CI 1214–1496	1355	Not reported	Unknown
51	GPT-5 (high)OpenAI7,036 votes · 55.5% win · 95% CI 1181–1447	1314	Not reported	Closed
52	DeepSeek V4 Pro 0813DeepSeek4,657 votes · 28.7% win · 95% CI 1170–1410	1290	Not reported	Open
53	Gemini 3 Flash PreviewGoogle5,548 votes · 37.5% win · 95% CI 1160–1418	1289	Not reported	Unknown
54	Gemini 2.5 ProGoogle7,000 votes · 49.7% win · 95% CI 1003–1575	1289	Not reported	Closed
55	Qwen 3.6 Plus PreviewAlibaba4,148 votes · 32.2% win · 95% CI 1136–1390	1263	Not reported	Unknown
56	Claude Sonnet 4.5 (32K Thinking)Anthropic7,709 votes · 47.6% win · 95% CI 1125–1369	1247	Not reported	Unknown
57	GPT-5 ProOpenAI6,553 votes · 53.0% win · 95% CI 1072–1326	1199	Not reported	Unknown

How to read this leaderboard

A higher rating means a model's text-prompt builds won more of their pairwise comparisons after Glicko-2 adjusted for opponent strength. Read the rating together with its confidence interval and vote count.

Operator receipt: 57 sourced rows are currently displayable on this page; the leading published row is GPT-6 Astra at 2724.

Honest limit: The score belongs to the full VoxelBench setup, including its generation tool, instructions, reasoning setting, prompt mix, renderer, and voter pool. It is not a controlled base-model spatial-reasoning score.

VoxelBench image-prompt leaderboard Multimodal and grounded rankings

Official live leaderboard Blind comparison voting

How we show VoxelBench text-prompt ratings

We mirror the text-prompt table from VoxelBench's public leaderboard API. The source page shows models after at least 50 votes and ranks the eligible rows with Glicko-2; the snapshot keeps each rating, deviation, 95% confidence interval, vote count, win rate, and win/loss/tie record.

Voters compare two anonymous voxel structures produced for the same text prompt. The result mixes model behavior with VoxelBench's generation tools, instructions, settings, prompt mix, renderer, and voter pool, so we keep it display only and separate from weighted model rankings.

Snapshot

57 eligible model rows50+ votes per modelText-prompt buildsBlind pairwise votesGlicko-2Display only

VoxelBench leaderboard Blind comparison voting VoxelBench prompts

The published VoxelBench Text snapshot places GPT-6 Astra first at 2724. The third row is 192 score units behind. The broader top-10 range is 618 score units, so the table still separates the published systems.

57 models have been evaluated on VoxelBench Text. The benchmark falls in the Multimodal & Grounded category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. VoxelBench Text is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About VoxelBench Text

Year

2025

Tasks

Live text prompts for 3D voxel construction

Format

Glicko-2 rating from blind pairwise votes

Difficulty

3D spatial construction and visual quality

We mirror the official text-prompt API rows that clear VoxelBench's 50-vote display gate. Glicko-2 ratings summarize blind pairwise preferences, while rating deviation and the 95% confidence interval show how uncertain each estimate remains.

VoxelBench leaderboard Public benchmark source

Freshness and provenance

Version

Live VoxelBench Glicko-2

Refresh cadence

Rolling

Staleness state

Current

Question availability

Prompts browsable; full task set not versioned

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does VoxelBench Text measure?

A live human-preference benchmark where language models turn text prompts into voxel structures and voters compare anonymous builds from the same prompt.

Which model leads the published VoxelBench Text snapshot?

GPT-6 Astra currently leads the published VoxelBench Text snapshot with 2724 glicko-2 rating. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on VoxelBench Text?

The November 12, 2026 snapshot contains 57 AI models.

Last updated: November 12, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.