AI Report Library

Stats

Prompt · Social media · food-buyers

“Where food buyers exist on social media”

Show full prompt

Find out where the food buyers exist on the social media? Output condition

  • Include fact links after the paragraph or sentence

Content density: websites cited vs. length in words

Takeaway and the numbers behind the dots

4 answers shown; 2 more left out because their source links were not captured.

Qwen3.8-Max is both the longest answer (793 words) and the one that cites the most websites (15).

Answer

Words

Websites

Websites per 1,000 words

Link

Sonnet 5.5High · Oct 6
Words394
Websites6
Per 1,000 words15.2
GPT-6.1 SolHigh · Oct 6
Words654
Websites12
Per 1,000 words18.3
Gemini 3.6Thinking · Oct 6
Words457
Websites6
Per 1,000 words13.1
Qwen3.8-MaxThinking · Oct 6
Words793
Websites15
Per 1,000 words18.9
Typical: the middle of each column
Words556
Websites9
Per 1,000 words16.8

Website overlaps by models

No website was cited by more than one model. Each model found its own sources.

Website overlaps by models: websites cited by more than one model
WebsiteSonnet 5.5GPT-6.1 SolGemini 3.6Qwen3.8-MaxCited by
Cited only by this model612615

A website counts once per model, however many of its pages are linked.

Length and websites vs. the typical answer

A model’s typical answer compared with the other models’ answers to the same prompts. Bars show the typical answer; each circle is one answer.

Model

Length vs. others

◂ shorter−50%avg+100%longer ▸

Websites vs. others

◂ fewer−50%avg+100%more ▸
Opus 5.54 answers on 4 prompts
Length
27% shorter
Websites
99% more
Sonnet 5.53 answers on 3 prompts
Length
about the same
Websites
about the same
GPT-6 Luna3 answers on 3 prompts
Length
38% shorter
Websites
about the same
GPT-6.1 Sol6 answers on 5 prompts
Length
34% longer
Websites
31% more
ChatGPT LatestOnly 1 answer so far
Length
90% longer
Websites
16% fewer
ChatGPT ProOnly 1 answer so far
Length
about the same
Websites
about the same
GPT-6 Astra3 answers on 3 prompts
Length
40% longer
Websites
59% fewer
Gemini 3.64 answers on 3 prompts
Length
40% longer
Websites
50% fewer
Grok 4.72 answers on 2 prompts
Length
51% shorter
Websites
73% fewer
Qwen3.8-Max2 answers on 2 prompts
Length
about the same
Websites
71% more

the model’s typical answereach answerbeyond the scale, with its numberThe shaded band is “about the same” (within 10%).

Left of the centre line is shorter or fewer, right is longer or more. avg is the middle answer from the other models on the same prompt, so a prompt only one model answered is left out. The scale is symmetric: an answer half as long as that sits as far left as one twice as long sits right (−50% and +100% are marked). One answer says little, so those rows are lighter. K3 is not shown: none of its answers had captured source links.

Based on 32 of 48 outputs on 8 prompts from 10 models, as of Oct 6, 2026. Facts, not rankings. How the numbers are counted
  • Left out: 16 outputs whose source links were not captured, which is why K3 does not appear. An output re-captured with its links joins automatically.
  • Words: the length of the report itself, counted from the first word of the report when a capture begins with tool commentary.
  • Websites: distinct site addresses (for example kaldi.co.jp, written without “www.”), each counted once per answer however many of its pages are linked.
  • Typical: the middle answer on a prompt. Half the answers are above it and half below.
  • Versus others: each answer is divided by the middle answer from the other models on the same prompt (marked “avg”). A model’s figure is the middle of those comparisons. “About the same” means within 10%.
  • Scale: circles sit on a symmetric scale from one fifth to five times that middle answer: half as long and twice as long are equally far from the centre. An answer beyond the scale is drawn at the edge as a dashed circle with its number.
  • Order: models always appear in the library’s fixed order. Nothing is sorted by a value, and nothing is scored.

About this library