BasedAGIBasedAGI
Menu
Rankings live

business_productivity

Document expansion

Expand short drafts into structured docs without changing meaning.

#1 Recommendation

gemini-2.5-pro

Strong on Galileo Agent Leaderboard v2 Avg TSQ (79%) and FACTS Benchmark Suite facts_grounding_score_pct (100%)

external/google/gemini-2-5-pro

24.8%

Score

42.4%

Confidence

Limited benchmark evidence for this use case.

52 ranked models with average evidence of 14.7 points. Rankings may shift as more benchmark data is ingested.

Ranked Models

30

Evidence Quality

81%

Scoring

Benchmark-backed

Top Signal

Galileo Agent Leaderboard v2: Avg TSQ

All Ranked Models

Max params:
Min confidence:
30 of 30
RankModelScore
#1gemini-2.5-pro

Strong on Galileo Agent Leaderboard v2 Avg TSQ (79%) and FACTS Benchmark Suite facts_grounding_score_pct (100%)

24.8%
#2gpt-4.1-20250414

Strong on OpenVLM OCRBench Official ocrbench_score_pct (88%) and Galileo Agent Leaderboard v2 Avg TSQ (64%)

24.6%
#3gemini-3-pro-preview

Strong on Vals Finance Agent overall_accuracy_pct (87%) and FACTS Benchmark Suite facts_grounding_score_pct (88%)

21.9%
#4gemini-2.5-flash
21.8%
#5gpt-5-mini-2025-08-07
19.6%
#6gpt-5-2025-08-07
18.7%
#7anthropic/claude-sonnet-4.6
17.9%
#8Grok-4-0709
17.3%
#9google/gemini-3.1-pro-preview
16.9%
#10openai/gpt-5.4-2026-03-05
16.4%
#12claude-sonnet-4-20250514
15.6%
#13gpt-4.1-mini-20250414
15.2%
#14gpt-5.1-2025-11-13
14.5%
#15claude-opus-4-5-20251101
13.9%
#16google/gemini-3.1-flash-lite-preview
13.6%
#18gpt-5.2-2025-12-11
12.9%
#19anthropic/claude-opus-4-6-thinking
12.5%
#20xai-org/grok-4-fast-reasoning
11.8%
#21gemini-3-flash-preview
11.8%
#22google/gemini-2.0-flash-001
11.7%
#23anthropic/claude-opus-4-5-20251101-thinking
11.7%
#27kimi/kimi-k2.5-thinking
11.1%
#28anthropic/claude-sonnet-4-5-20250929-thinking
11.0%
#29xai-org/grok-4-1-fast-reasoning
11.0%
#35gpt-4o
9.9%
#37anthropic/claude-opus-4-1-20250805
9.7%
#41anthropic/claude-haiku-4-5-20251001-thinking
9.3%
#43mistralai/mistral-large-2512
9.3%
#45gpt-4o-2024-08-06
9.2%
#47alibaba/qwen3.5-flash
9.1%

Compare Models

Model A leads by +0.2%

Shareable Link →

Model A

gemini-2.5-pro

external/google/gemini-2-5-pro

24.8%

Rank #1

Confidence 42.4%27 evidence pts

Galileo Agent Leaderboard v2: Avg TSQ

Value 79.5% · Conf 100.0% · Weight 2.4%

galileo_agent_v2.avg_tsq (Mar 12, 2026)

FACTS Benchmark Suite: facts_grounding_score_pct

Value 100.0% · Conf 100.0% · Weight 2.0%

facts_benchmark_suite.facts_grounding_score_pct (Mar 12, 2026)

OpenVLM OCRBench Official: ocrbench_score_pct

Value 90.7% · Conf 100.0% · Weight 2.0%

openvlm_ocrbench_official.ocrbench_score_pct (Mar 12, 2026)

Vals Mortgage Tax: overall_accuracy_pct

Value 99.0% · Conf 100.0% · Weight 1.7%

vals_mortgage_tax.overall_accuracy_pct (Mar 12, 2026)

Model B

gpt-4.1-20250414

external/openai/gpt-4-1-20250414

24.6%

Rank #2

Confidence 37.3%25 evidence pts

OpenVLM OCRBench Official: ocrbench_score_pct

Value 87.7% · Conf 100.0% · Weight 1.9%

openvlm_ocrbench_official.ocrbench_score_pct (Mar 12, 2026)

Galileo Agent Leaderboard v2: Avg TSQ

Value 64.1% · Conf 100.0% · Weight 1.9%

galileo_agent_v2.avg_tsq (Mar 12, 2026)

Vectara HHEM Leaderboard: overall_hallucination_error_pct

Value 82.5% · Conf 100.0% · Weight 1.8%

vectara_hhem_leaderboard.overall_hallucination_error_pct (Mar 12, 2026)

OpenVLM TextVQA Official: textvqa_score_pct

Value 76.8% · Conf 100.0% · Weight 1.7%

openvlm_textvqa_official.textvqa_score_pct (Mar 12, 2026)

Ranking Diagnostics & Missing Models

Source Lift

Ranked

52

Sources

8

Quality

Insufficient

Vals MedQA

vals_medqa

37 rows

0.4% avg lift

Vals Legal Bench

vals_legal_bench

36 rows

0.4% avg lift

Vals Tax Eval v2

vals_tax_eval_v2

35 rows

0.4% avg lift

Vals Mortgage Tax

vals_mortgage_tax

33 rows

1.4% avg lift

Missing Strong Models

gpt-4o-20241120

external/openai/gpt-4o-20241120

Rank #49

10.7%

Thin evidence after weighting
Taxonomy Details

Core Tasks

task.rewrite_expansiontask.outline_generation

Required Modes

none

Domains

domain.general_business

Related Use Cases