Design Arena Elo

Design Arena gives the same prompt to several models and shows voters two of the outputs side by side, without model names. Each choice is one pairwise vote. The votes are turned into a rating with a Bradley-Terry model, shown as an Elo-style number that starts at 1200, with a separate leaderboard per category such as website, UI component, SVG, data visualization, 3D and game development. A rating reflects human preference on visual and interactive output. It depends on who votes and which prompts are used, and ratings from different categories are not on the same scale.

Sources: Design Arena: how the rankings work

Open-weight models

Design Arena · data from 2026-10-11
ModelCategoryEndpointWin rate
Kimi K33dmoonshotai/kimi-k3-20260715139467.7%
GLM 5.33dz-ai/glm-5.3-20260816134560.7%
GLM 5.3 Flash3dz-ai/glm-5.3-flash-20260826132156.5%
DeepSeek V4 Pro3ddeepseek/deepseek-v4-pro-20260423125756.8%
Kimi K2.7 Code3dmoonshotai/kimi-k2.7-code-20260612124550.5%
MiniMax M33dminimax/minimax-m3-20260531121250.2%
Nemotron 3 Ultra3dnvidia/nemotron-3-ultra-550b-a55b-20260604113539.2%
Step 3.7 Flash3dstepfun/step-3.7-flash-20260528113240.6%
GLM 5.3 Flashasciiartz-ai/glm-5.3-flash-20260826127255.0%
GLM 5.3asciiartz-ai/glm-5.3-20260816120746.1%
Kimi K2.7 Codeasciiartmoonshotai/kimi-k2.7-code-20260612120548.6%
MiniMax M3asciiartminimax/minimax-m3-20260531116946.8%
Step 3.7 Flashasciiartstepfun/step-3.7-flash-20260528115844.9%
DeepSeek V4 Proasciiartdeepseek/deepseek-v4-pro-20260423115546.4%
Nemotron 3 Ultraasciiartnvidia/nemotron-3-ultra-550b-a55b-20260604110438.1%
Kimi K3codecategoriesmoonshotai/kimi-k3-20260715136762.8%
GLM 5.3codecategoriesz-ai/glm-5.3-20260816131254.7%
GLM 5.3 Flashcodecategoriesz-ai/glm-5.3-flash-20260826128749.4%
Kimi K2.7 Codecodecategoriesmoonshotai/kimi-k2.7-code-20260612125750.0%
MiniMax M3codecategoriesminimax/minimax-m3-20260531125250.9%
DeepSeek V4 Procodecategoriesdeepseek/deepseek-v4-pro-20260423124451.9%
Step 3.7 Flashcodecategoriesstepfun/step-3.7-flash-20260528118642.6%
Nemotron 3 Ultracodecategoriesnvidia/nemotron-3-ultra-550b-a55b-20260604114936.1%
Kimi K3datavizmoonshotai/kimi-k3-20260715135463.2%
GLM 5.3datavizz-ai/glm-5.3-20260816126952.7%
GLM 5.3 Flashdatavizz-ai/glm-5.3-flash-20260826126650.3%
MiniMax M3datavizminimax/minimax-m3-20260531123951.3%
Kimi K2.7 Codedatavizmoonshotai/kimi-k2.7-code-20260612123148.8%
DeepSeek V4 Prodatavizdeepseek/deepseek-v4-pro-20260423120648.7%
Step 3.7 Flashdatavizstepfun/step-3.7-flash-20260528117643.0%
Nemotron 3 Ultradataviznvidia/nemotron-3-ultra-550b-a55b-20260604114437.4%
Kimi K3gamedevmoonshotai/kimi-k3-20260715136961.4%
GLM 5.3gamedevz-ai/glm-5.3-20260816132052.8%
GLM 5.3 Flashgamedevz-ai/glm-5.3-flash-20260826128847.2%
DeepSeek V4 Progamedevdeepseek/deepseek-v4-pro-20260423123652.3%
Kimi K2.7 Codegamedevmoonshotai/kimi-k2.7-code-20260612122547.0%
MiniMax M3gamedevminimax/minimax-m3-20260531121846.0%
Step 3.7 Flashgamedevstepfun/step-3.7-flash-20260528116538.9%
Nemotron 3 Ultragamedevnvidia/nemotron-3-ultra-550b-a55b-20260604114436.4%
Kimi K3svgmoonshotai/kimi-k3-20260715130559.8%
GLM 5.3svgz-ai/glm-5.3-20260816129456.8%
GLM 5.3 Flashsvgz-ai/glm-5.3-flash-20260826128857.1%
Kimi K2.7 Codesvgmoonshotai/kimi-k2.7-code-20260612117544.4%
MiniMax M3svgminimax/minimax-m3-20260531116245.2%
DeepSeek V4 Prosvgdeepseek/deepseek-v4-pro-20260423113945.4%
Step 3.7 Flashsvgstepfun/step-3.7-flash-20260528108137.2%
Nemotron 3 Ultrasvgnvidia/nemotron-3-ultra-550b-a55b-20260604106832.5%
Kimi K3uicomponentmoonshotai/kimi-k3-20260715135661.4%
GLM 5.3 Flashuicomponentz-ai/glm-5.3-flash-20260826132455.0%
GLM 5.3uicomponentz-ai/glm-5.3-20260816132157.4%
Kimi K2.7 Codeuicomponentmoonshotai/kimi-k2.7-code-20260612126551.0%
MiniMax M3uicomponentminimax/minimax-m3-20260531125050.9%
DeepSeek V4 Prouicomponentdeepseek/deepseek-v4-pro-20260423122550.6%
Step 3.7 Flashuicomponentstepfun/step-3.7-flash-20260528118542.2%
Nemotron 3 Ultrauicomponentnvidia/nemotron-3-ultra-550b-a55b-20260604114536.4%
Kimi K3websitemoonshotai/kimi-k3-20260715133959.5%
GLM 5.3websitez-ai/glm-5.3-20260816130354.6%
GLM 5.3 Flashwebsitez-ai/glm-5.3-flash-20260826127648.5%
Kimi K2.7 Codewebsitemoonshotai/kimi-k2.7-code-20260612126851.4%
MiniMax M3websiteminimax/minimax-m3-20260531126251.9%
DeepSeek V4 Prowebsitedeepseek/deepseek-v4-pro-20260423123950.5%
Step 3.7 Flashwebsitestepfun/step-3.7-flash-20260528119944.2%
Nemotron 3 Ultrawebsitenvidia/nemotron-3-ultra-550b-a55b-20260604114534.9%

Measured by Design Arena on hosted endpoints, not by this site. A quantized file you run yourself may score differently. One model can have several endpoints. Source: Design Arena (www.designarena.ai) via OpenRouter (openrouter.ai/rankings). designarena.ai