Models ranked from best to worst on average of all tasks
Writing instructions adherence
Stylistic and vocabulary diversity
Self-evaluation capability
The amount of time to onboard someone for content marketing, back and forth, feedback.
o3-mini
 94.7%
89.8%
         2.024
81.3%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Claude Sonnet 3.5 (Oct. 2024 model)
      91.6%
89.6%
2.103
         78.7%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Qwen 2.5-72B
      90.6%
         80.8%
         1.988
84.5%
The amount of time to onboard someone for content marketing, back and forth, feedback.
GPT 4o (Nov. 2024 model)
      92.0%
         75.4%
2.083
         79.4%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Claude Sonnet 3.7
      91.9%
         86.4%
         1.932
         76.8%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Gemini 2.0 Flash-Lite
 92.5%
         86.0%
1.900
         74.5%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Qwen 2.5-7B
      89.9%
89.8%
         1.992
44.9%
The amount of time to onboard someone for content marketing, back and forth, feedback.
GPT 4o-mini
 92.4%
         77.4%
         2.055
39.0%
The amount of time to onboard someone for content marketing, back and forth, feedback.
DeepSeek V3
      90.1%
         81.2%
         2.024
         78.3%
The amount of time to onboard someone for content marketing, back and forth, feedback.
GPT 4o (Aug. 2024 model)
 84.2%
54.6%
         2.079
         80.9%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Llama 3.1-405B
      85.6%
         80.3%
         2.021
         79.0%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Mistral Small 3
 75.6%
         76.6%
         2.046
         80.7%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Gemini 1.5 Pro
      89.3%
71.4%
         1.988
         80.6%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Gemini 2.0 Flash
      86.5%
         81.6%
         1.980
         71.4%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Claude Haiku 3.5
      92.1%
         77.8%
1.919
         69.2%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Gemini 1.5 Flash
      86.9%
         73.2%
1.927
80.9%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Llama 3.3-70B
      88.1%
         80.6%
         1.938
         65.5%
The amount of time to onboard someone for content marketing, back and forth, feedback.
Llama 3.1-8B
 82.4%
64.8%
2.086
50.4%

Main author: Azamat Omuraliev
Collaborators: Sam Baud, Sjoerd de Kreij, Jirmaine Tan, Mayuran Visakan, Grzegorz Lazka, Yun Yang, Jeroen Meij

Read about the methodology in our blogpost

Risk free try out

🚀 Discover how Sarah can transform your marketing in just 15 minutes 🔥 Limited slots available - Don't miss out!

Grow on autopilot, get unburdened by Sarah.