supermarket

Benchmark

Deep Research Bench (FutureSearch)

Measures: 91 web research tasks of eight types (finding a number, compiling a dataset, gathering evidence, validating a claim and others) that an agent answers by searching and reading pages.

Maintainer
FutureSearch
Introduced
2025
Status
Active
Saturation
Headroom
Typical provenance
Independent
In scoring set
Yes

Official leaderboard

Leaderboard

Show independent results only

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

Leaderboard
ModelAverage scoreProvenanceSourceEvaluated
Claude Opus 4.6
Anthropic · High effort
55.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Sonnet 4.6
Anthropic · High effort
54.9%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.5
Anthropic · High effort
54.8%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5.5
OpenAI · High effort
54.0%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.5
Anthropic · Low effort
53.7%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.6
Anthropic · Medium effort
53.2%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Sonnet 4.5
Anthropic · 2K effort
52.6%Third partyDeepResearchBench Leaderboard via Epoch AI—
Claude Opus 4.6
Anthropic · Low effort
51.4%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Sonnet 4.6
Anthropic · Low effort
50.4%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.8
Anthropic · High effort
50.2%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3 Flash
Google DeepMind · Low effort
49.8%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5
OpenAI · Low effort
49.6%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5.5
OpenAI · Medium effort
49.6%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.8
Anthropic · Low effort
49.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3 Flash
Google DeepMind · Minimal effort
49.0%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5
OpenAI · Minimal effort
48.9%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5.5
OpenAI · Low effort
48.7%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5
OpenAI · Medium effort
48.6%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.1
Anthropic
48.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5
OpenAI · High effort
48.1%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3 Flash
Google DeepMind · High effort
47.9%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3.1 Pro
Google DeepMind · High effort
47.8%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Sonnet 4.5
Anthropic · Low effort
47.5%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4.8
Anthropic · Medium effort
47.4%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Grok 4
SpaceXAI (xAI)
47.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Opus 4
Anthropic · 2K effort
46.8%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Sonnet 4
Anthropic · 2K effort
46.6%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3 Pro
Google DeepMind · Low effort
46.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude Haiku 4.5
Anthropic · Low effort
45.5%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
o3
OpenAI · Medium effort
45.2%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3.1 Pro
Google DeepMind · Low effort
44.5%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Claude 3.7 Sonnet
Anthropic · 2K effort
43.6%Third partyDeepResearchBench Leaderboard via Epoch AI—
Gemini 2.5 Pro
Google DeepMind
42.8%Third partyDeepResearchBench Leaderboard via Epoch AI—
GPT-5.1
OpenAI · Low effort
42.8%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 2.5 Pro
Google DeepMind
41.5%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5.2
OpenAI · Low effort
41.1%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3.1 Flash-Lite
Google DeepMind · Low effort
37.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
Gemini 3.1 Flash-Lite
Google DeepMind
36.4%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
GPT-5.4 mini
OpenAI · Low effort
36.3%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
DeepSeek-R1
DeepSeek
35.1%Third partyDeepResearchBench Leaderboard via Epoch AI—
GPT-5.4
OpenAI · Low effort
35.1%Third partyhttps://evals.futuresearch.ai/#drb via Epoch AI—
  • Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.

Sources