Kimi K3 Outperforms Inkling on AI Benchmarks Despite Larger Size
Reactions from ranked influencers
2 posts
a couple folks asked how kimi k3 compares to thinking machines' inkling table / charts below! (note: k3 is 2.8T parameters and Inkling is 975B, so not an apples-to-apples size comparison!)
k3 leads on all our benchmarks: 💊 Antidote (everyday chat): K3 1081 Elo, Inkling 986 📋 HANDBOOK.md (long-context agents): 11.9% vs 1.9% 📊 Chartography (chart understanding): 26.6% vs 3.4% 🧮 Riemann-bench (research math): 37.6% vs 15.2% 🧩 ComplexConstraints (enterprise IF): 37.9% vs 0.3% inkling's an early release, we'll rerun as new checkpoints land! http://surgehq.ai/benchmarks