Moonshot’s Kimi K3 tops DesignArena with a 1408 Elo and unusually long reasoning traces
cephaloform · x · 2026-07-24
Moonshot’s Kimi K3 is drawing attention after topping DesignArena’s single-shot Frontend Arena with an Elo of 1408, according to the cited article.
- The post highlights a massive increase in consumed thinking tokens during coding-style reasoning.
- It claims Kimi K3 uses about 10× more thinking tokens for coding than any other model observed.
- The model is also said to have effectively memorized Unsplash image IDs, enabling stronger image reasoning without a web search tool call.
- The attached visual shows a very long thinking trace broken into steps like planning, choosing images, deciding page sections, animation design, CSS design, header design, checkout design, and micro-interaction design.
- The result reportedly helps generate websites with higher relevance and accuracy on DesignArena than previously seen.
Related event: Kimi K3 Tops Code and Design Arenas Amid Distillation Debate(7 posts)→
More from Models
- DeepSeek V4.1 Flash Hits 98% of GPT-6 Astra's Score at 1.4% of the Cost in Third-Party Benchmark — ayushtweetshere · 2026-09-11
- TheZvi Polls: Has Your Coding Model Choice Changed Since Fable 5.1 and Astra? — TheZvi · 2026-09-11
- antirez Weighs In on Anthropic Banning Minors From Using Claude — antirez · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- Fully local voice assistant on an RTX 3060 replicates the GPT Live demo in 6.5 minutes — liampetti · 2026-09-11
- Meta's Muse Agent has built-in invite code logic, hinting at free-usage expansion — testingcatalog · 2026-09-11