Suspected Kimi K3 Spotted in the Arena
testingcatalog · x · 2026-07-15
A new model named Kivine has appeared on @arena, which is highly likely an early version of Moonshot/Kimi K3.
The author shares some comparative observations:
- In a Universe simulation prompt, Fable 5 finishes faster and has more stable UI components.
- Kimi K3 delivers a more complex output that is visually more engaging.
- During certain interactions, like selecting a planet and cranking the speed to x100, the model generates an exaggerated first-person rotation effect.
Overall verdict: the two are very close in capability, but they differ in terms of interaction stability and visual complexity.
Related event: Kimi K3 hype builds as KIVINE appears on Arena(43 posts)→
More from Models
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11