Cognition's FrontierCode 2.0 chart accused of cherry-picking after DeepSeek score gaps surface
teortaxesTex · x · 2026-10-29
A widely shared Cognition chart comparing Terminal-Bench 4.0 vs FrontierCode 2.0 is under fire for apparent cherry-picking.
- DeepSeek V4.1 Flash scores 77.0% on DeepSWE v1.1 but only 19.3% on FrontierCode 2.0
- Its near-identical scores on TB 4.0 (self-reported) and FC 1.1 (AA-verified) raise further doubts
- The 77 vs 74 presentation is called misleading — another episode in vendor-selected-benchmark skepticism
More from Models
- Dev Warns Google Carbon Demand Would Collapse Without Argon's $10 Discount Pricing — bindureddy · 2026-10-29
- Gemini 4 Carbon Tops Artificial Analysis Intelligence Index at 59, No Public API Yet — kimmonismus · 2026-10-29
- Google ships Gemini 4 Carbon two weeks after Argon, topping Artificial Analysis V5 — BenBlaiszik · 2026-10-29
- MiniMax M3.1 Flash scores 43 on Artificial Analysis index as skeptics predict flop — zacharynado · 2026-10-29
- Andon Labs: AI Now Ties Humans on Physical 3D Understanding Benchmarks — ResultBackground2450 · 2026-10-29
- LMArena launches an arena-style benchmark for agent capabilities — allenainie · 2026-10-29