AIMoonshot AI's Kimi-K3 tops Frontend Code Arena
It matches Claude Fable 5 performance at 35% cost.
Combined views
8.3M
118 Sources, first seen 67d ago
Moonshot AI's Kimi-K3 tops Frontend Code Arena
It matches Claude Fable 5 performance at 35% cost.
Sources
AM 
Amjad Masad@amasadApparently the distillation model can outperform the teacher model 😂 https://twitter.com/arena/status/2077824029126504525
- likes: 4.3K
- replies: 178
- bookmarks: 492
- reposts: 274
BR Bindu Reddy@bindureddyKimi K3 Is The Best Open-Source Model In The World - scores right below Opus 4.8 - works very well on first turn benchmark style questions - beats GLM 5.2 Try it on ChatLLM now
- likes: 11
- replies: 0
- bookmarks: 3
- reposts: 2
WH wh@nrehiew_This is the most important plot of all the benchmarks. Cost per intelligence, on average, is just under the Pareto Frontier set by GPT 5.6 Sol, significantly better than Opus or even Fable. Hopefully, someone makes it run faster/cheaper + distills it…
- likes: 70
- replies: 7
- bookmarks: 16
- reposts: 3
PW Peter Wildeford🇺🇸🚀@peterwildefordThe Kimi K3 benchmarks seem impressive, but I'd like to wait for more third-party confirmation before coming to strong judgements about geopolitics. https://twitter.com/zephyr_z9/status/2077805978268135480
- likes: 15
- replies: 0
- bookmarks: 0
- reposts: 1
KD Kosta Derpanis (sabbatical in Zurich)@CSProfKGDIsn’t competition great? https://twitter.com/zephyr_z9/status/2077805978268135480
- likes: 7
- replies: 0
- bookmarks: 0
- reposts: 0
BR Bindu Reddy@bindureddyKIMI K3 CLOSES THE GAP BUT RANKS BEHIND FRONTIER MODELS Our benchmark, LiveBench, has a lot of hidden questions and models can't memorize them K3 is the best good open-source model but is below Opus 4.8, Sol and Fable Also in practice, Kimi spins a lot and costs as much as…
- likes: 131
- replies: 29
- bookmarks: 26
- reposts: 12