Positive users congratulated OpenAI on GPT-5.6 Sol topping the Design Arena with a record Elo rating and faster, better results, while negative users rejected the benchmark rankings as flawed and accused the company of misconduct.
Based on 85 visible X reactions from 155 accounts.
Ask a question below.
Published answers will appear here.
A request for ChatGPT rooms: please let each conversation retain its chosen model unless the user manually changes that specific room. When I switch models in one room, it should not silently change every other room. For users who rely on continuity, tone, memory, and established workflows, the model is part of the room’s identity. Please add a “lock model for this conversation” option.
@rohanpaul_ai Exciting to see measurable gains in real-world design quality. 🚀 The pace of innovation in AI is simply remarkable.
@gdb That's awesome. Congrats on hitting it
@gdb congrats!
GPT-5.6 Sol just took first place on Design Arena, reaching 1353 Elo in frontend design and above Claude Fable 5. Its 60-point jump over GPT-5.5 moved OpenAI upward by 18 leaderboard positions. Design Arena asks models to create webpages from prompts without an agentic tool loop. Human voters compare anonymized outputs, and pairwise preferences become Elo scores through statistical modeling. A higher score therefore reflects preferred visual results, not guaranteed software correctness or reliability. GPT-5.6 Sol's superior ability in inspecting rendered-page likely explains why its design quality improved without requiring longer interactions. Its speed adds another advantage, because comparable rivals need more time for similar preference scores.
The model scored 1353 Elo, beating its predecessor by 60 points.
A request for ChatGPT rooms: please let each conversation retain its chosen model unless the user manually changes that specific room. When I switch models in one room, it should not silently change every other room. For users who rely on continuity, tone, memory, and established workflows, the model is part of the room’s identity. Please add a “lock model for this conversation” option.
@rohanpaul_ai Exciting to see measurable gains in real-world design quality. 🚀 The pace of innovation in AI is simply remarkable.
GPT-5.6 Sol just took first place on Design Arena, reaching 1353 Elo in frontend design and above Claude Fable 5. Its 60-point jump over GPT-5.5 moved OpenAI upward by 18 leaderboard positions. Design Arena asks models to create webpages from prompts without an agentic tool loop. Human voters compare anonymized outputs, and pairwise preferences become Elo scores through statistical modeling. A higher score therefore reflects preferred visual results, not guaranteed software correctness or reliability. GPT-5.6 Sol's superior ability in inspecting rendered-page likely explains why its design quality improved without requiring longer interactions. Its speed adds another advantage, because comparable rivals need more time for similar preference scores.
And this doesn't even include the cost dimension! https://twitter.com/Designarena/status/2076391367446860249
finally catching up to GLM 5.2 https://twitter.com/designarena/status/2076391367446860249
big milestone! https://twitter.com/designarena/status/2076391367446860249
Positive users congratulated OpenAI on GPT-5.6 Sol topping the Design Arena with a record Elo rating and faster, better results, while negative users rejected the benchmark rankings as flawed and accused the company of misconduct.
Based on 85 visible X reactions from 155 accounts.
Ask a question below.
Published answers will appear here.
And this doesn't even include the cost dimension! https://twitter.com/Designarena/status/2076391367446860249