Benchmark profile
VoxelBench Text-Prompt Leaderboard (VoxelBench Text)
A live human-preference benchmark where language models turn text prompts into voxel structures and voters compare anonymous builds from the same prompt.
Data verifiedHow 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: 49 sourced rows are currently displayable on this page; the leading published row is Claude Opus 5 (Max) at 2236.
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.
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.
Glicko-2 rating on VoxelBench text-prompt leaderboard — July 29, 2026
BenchLM mirrors the published glicko-2 rating view for VoxelBench text-prompt leaderboard. Claude Opus 5 (Max) leads the public snapshot at 2236 , followed by GPT-5.6 Sol (Max) (2235) and Claude Fable 5 (Max) (2191). BenchLM does not use these results to rank models overall.
Claude Opus 5 (Max)
Anthropic
859 votes · 90.3% win · 95% CI 2156–2316
claude-opus-5
GPT-5.6 Sol (Max)
OpenAI
2,516 votes · 93.0% win · 95% CI 2143–2327
gpt-5.6-sol
Claude Fable 5 (Max)
Anthropic
3,499 votes · 91.4% win · 95% CI 2097–2285
claude-fable-5
Glicko-2 rating table (49 models)
ScoreThe published VoxelBench Text snapshot places Claude Opus 5 (Max) first at 2236. The third row is 45 score units behind. The broader top-10 range is 550 score units, so the table still separates the published systems.
49 models have been evaluated on VoxelBench Text. The benchmark falls in the External benchmark mirrors 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.
BenchLM freshness & provenance
Version
Live VoxelBench Glicko-2
Refresh cadence
Rolling
Staleness state
Current
Question availability
Prompts browsable; full task set not versioned
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.
FAQ
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?
Claude Opus 5 (Max) currently leads the published VoxelBench Text snapshot with 2236 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?
49 AI models are included in BenchLM's mirrored VoxelBench Text snapshot, based on the public leaderboard captured on July 29, 2026.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.