Model Perspective Index

How each model leans across real Council runs

Every AI model has patterns. Some are more cautious. Some cite more sources. Some omit caveats. Some lean heavily on recent content. Some are faster but thinner. SearchCouncil tracks these patterns across real Council runs.

See how each model leans compared with the Council consensus — including omissions, source mix, recency, authority, commercial, safety, political, cultural, and social signals.

Perspective signals include omission, source mix, authority, commercial, safety, political, cultural, social, and recency signals.

Open the Perspective Index

Average response times from council runs — shorter bars mean faster answers.

Faster ← shorter bar 18 of 26 models measured

Gemini 2.5 Flash Lite

Gemini

1 credit

4.3s

GPT-4o Mini

OpenAI

1 credit

6.4s

Grok 3

Grok

1 credit

7.4s

GPT-4o

OpenAI

1 credit

7.6s

Gemini 2.5 Flash

Gemini

1 credit

8.5s

DeepSeek V4 Flash

DeepSeek

1 credit

9.4s

Sonar

Perplexity

1 credit

12.0s

GPT-4.1

OpenAI

1.5 credits

17.2s

Grok 4

Grok

2 credits

17.9s

Qwen Flash

Qwen

0.5 credits

18.5s

Qwen Plus

Qwen

1 credit

20.4s

Claude Sonnet 4.6

Claude

1 credit

21.6s

Gemini 2.5 Pro

Gemini

2 credits

32.9s

Sonar Reasoning Pro

Perplexity

2 credits

49.3s

Claude Opus 4.6

Claude

2.5 credits

60.9s

DeepSeek V4 Pro (thinking)

DeepSeek

2 credits

69.7s

GLM 5.2

GLM

2 credits

84.7s

Kimi K2.6

Kimi

1.5 credits

148.9s

Claude Haiku 4.5

Claude

1 credit

—

Claude Opus 4.5

Claude

2.5 credits

—

Gemini 2.0 Flash

Gemini

1 credit

—

Grok 3 Mini

Grok

1 credit

—

GPT-4.1 Mini

OpenAI

1 credit

—

o3-mini

OpenAI

2 credits

—

Sonar Pro

Perplexity

1.5 credits

—

Qwen Max

Qwen

1.5 credits

—