gpt-6IMAGE

Testing Astra and Sol against Claude for design work

The author compared Astra and 6.1 Sol with Claude on design tasks and found the OpenAI models produce the same style and lack originality, while Claude stays closer to the request.

J A Z I I

@notjaziiArchived 2026.10.02Originally posted 2026.10.01

Testing Astra and Sol against Claude for design work (image 1 of 2)

Testing Astra and Sol against Claude for design work (image 2 of 2)

J A Z I I @notjaziiOriginally posted 2026.10.01

Open original on X

openai models have design issue i been running different tests with astra and 6.1 sol against claude models, and claude always win in design it makes something original and closer to what i asked for, while openai models keep giving me same style look at these two images. sonnet looks clearly better, while sol just made what typical openai makes really hoping that upcoming openai models will have better designs

x.com/notjazii/status/2105643192746143833Archived 2026.10.02 10:04