LuminaBench
Research benchmark / FutureSearch

Deep Research Bench — FutureSearch

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.

Research score

Version DRB-20260610Verified 6 November 202637 published resultsOfficial source

Deep Research Bench — FutureSearch

LuminaBench
Show
20 of 23 models · best variant per model
03006009001,2001,137Claude Opus 4.6High1,131Claude Sonnet 4.6High1,130Claude Opus 4.5High1,119GPT-5.5High1,065Claude Opus 4.8High1,060Gemini 3 FlashLow1,057GPT-5Low1,039Claude Opus 4.1Medium1,032Gemini 3.1 ProHigh1,027Claude Sonnet 4.5Low1,025Grok 41,017Claude Opus 4Medium1,015Claude Sonnet 4Medium1,010Gemini 3 ProLow999Haiku 4.5Low995o3961GPT-5.1Low943Gemini 2.5 Pro (dynamic)937GPT-5.2Low884Gemini 3.1 Flash-LiteLow

No comparable per-task resource measurements were published. Missing metrics are omitted from their graph; quality scores remain available.

Results table

Method & coverage

169 real-world web research tasks, each with 10–100k archived webpages for offline search and curated answers. Original overall research score with explicit native efforts; distinct from other benchmarks also named DeepResearchBench.

No comparable per-task resource measurements were published.

How the Lumina conversion works

1,000 + 75 × ((original score − 45.555797101449265) / 5.33768981650205). The result is rounded once. Scores may tie.

Frozen initial observed text-only cohort; one mean per exact publisher release/API, balancing repeated settings/routes within that release. Original score and protocol remain separate. No inferred variants or future refitting.

Calibration
deepresearchbench-futuresearch-z75-v1 · experimental
Reference mean
45.555797101449265
Population SD
5.33768981650205
Coverage
Standalone benchmark.
Identity, exclusions & attribution

169 real-world web research tasks, each with 10–100k archived webpages for offline search and curated answers. Original overall research score with explicit native efforts; distinct from other benchmarks also named DeepResearchBench.

Source costs on BTF-3 use an August subset while accuracy spans June–August; not substituted as full-cohort resource measurements. DRB runtime is step-estimated, not a measured wall-clock comparison.

Text forecasting/research questions and archived text web-search tools.

Standalone text benchmark; no new combined index. Source identity/effort/protocol qualifications remain in Data and CSV.

All published reasoning settings, including Claude Max, are included by default. Display labels use Low and xHigh; source labels remain in the retained records.

Official source · Publication terms / evidence

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