Kimi K3 ECI 157 Beats V4-Pro; DeepSeek V4-Flash 153 Second Open Model
Jsevillamol · x · 2026-08-05
Epoch AI's ECI shows Kimi K3 at 157 (range 155-161), surpassing V4-Pro-preview's 149 and V4-Flash's 153. Sol 5.6 scores 162. On FrontierMath tier 4, Kimi K3 hits 24%, matching Grok 4.5, while V4-Pro-preview only 2%. DeepSeek V4-Flash-0731 with ECI 153 is second open model.
More from Models
- GPT 5.6 Sol Successfully Formalizes Complex Math Proof for Nonsofic Existence — Sauers_ · 2026-08-05
- GPT 5.6 Successfully Formalizes Complex Math Proofs Without Internal Lean Code — Sauers_ · 2026-08-05
- Chinese Models Dominate Forecasting Leaderboard with Advanced AI Agents — teortaxesTex · 2026-08-05
- Developer Take: Free Gemini 2.0 Flash Offers Better Value Than Kimi — mertdumenci · 2026-08-05
- Quantized DeepSeek V3 hits 247 tokens/s decode in just 162GB — teortaxesTex · 2026-08-05
- Nous Research Co-founder: Model Sycophancy is a 'Reward Hack', Not Loyalty — petergyang · 2026-08-05