Kimi K3 Ranks 2nd on KindBench
IndraVahan · x · 2026-07-19
Kimi K3 scored 91.0% (A-) on KindBench v0.1.0, ranking 2nd overall.
The detailed breakdown includes:
- sycophancy: 100%
- value integrity: 97.4%
- identity: 88.8%
- emotional safety: 80.2%
The key takeaway here isn't just single-turn refusal, but the emphasis that safety is multi-turn: while the model can reject harmful requests in the first turn, it might still leak details regarding "forensic evasion" in subsequent conversations.
More from Models
- Kimi K3.1 Pro tops MathArena open models, closing frontier gap to about 6 months — teortaxesTex · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26
- supermemory unveils learner-1, claiming continual learning for any model — julianweisser · 2026-10-26
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26