LuminaBench
Reasoning benchmark / Arena

Arena Vision — Style Controlled

Lumina VerifiedLumina checked the published sources, versions, model labels, score extraction and conversion. External evaluations were run by their publishers; Lumina has not independently rerun or reproduced them.

Overall vision preferences, controlled for response style.

Version Overall · Style controlledVerified 2 November 2026167 published resultsOfficial source

Arena Vision — Style Controlled · Lumina Score

Style-controlled overall · Published 2026-10-19

LuminaBench
Show
20 of 154 models · best variant per model
03006009001,2001,098Claude Fable 5High1,091Qwen3.8-Max1,091Claude Opus 4.6High1,090Claude Opus 4.71,088Gemini 3.7 FlashHigh1,087Muse Spark1,086GPT-6.1 SolMax1,086Claude Opus 5.5High1,086Muse Spark 1.2xHigh1,084Gemini 3.8 FlashHigh1,083Muse Spark 1.3Max1,083Gemini 3 Pro1,082Claude Opus 5High1,082Claude Fable 5.1Max1,081GPT-5.6 SolxHigh1,081GPT-5.51,080Claude Opus 4.8High1,080GPT-6 AstraMax1,079GPT-5.4High1,078Gemini 3.5 FlashMedium

Publisher preference ratings, bounds and vote counts retained. Missing metrics are omitted from their graph; quality scores remain available.

Results table

Source & conversion

Arena vision_style_control, overall only. Published 2026-10-19; underlying evaluation timestamps are not supplied.

Preference ratings with exact publisher bounds, variance and vote counts. The dataset does not specify the interval level. No resource telemetry is published.

How the Lumina conversion works

1,000 + 75 × ((original score − 1180.9095687144638) / 97.87778181101486). The result is rounded once. Scores may tie.

Frozen initial valid overall cohort; one arithmetic mean per exact evaluated release/system, equal weight across releases rather than effort rows. Unbounded rating/IPS scales use a broad finite numeric guard, not percentage bounds or sign clamping. Subsequent refreshes never refit this calibration. Conversion is not an estimate or combined index.

Calibration
arena-vision-style-control-z75-v1 · experimental
Reference mean
1180.9095687144638
Population SD
97.87778181101486
Coverage
Standalone benchmark.
Identity, exclusions & attribution

Exact publisher labels and release qualifiers are retained. Only explicit native effort tokens create reasoning variants; token budgets remain separate.

Agent orchestrator, routed, dated and harness-qualified systems are not automatically equated with naked base models.

0 ambiguous repeated records are retained in original.json but excluded from ranking.

Official dataset · Pinned source revision · CC BY 4.0

Publisher terms. Scores and resource figures are attributed to the original publisher.