DeepSeek V4F Beats V4P at Chess, Writes Its Own Engine Mid-Game
teortaxesTex · x · 2026-08-03
A user pitted DeepSeek V4F GA against V4P Preview in a game of chess, with the white-playing V4F GA achieving a decisive victory.
The highlight of the match was the distinct behavioral differences between the models: V4F GA actually wrote its own chess engine mid-game to assist with calculations and made zero illegal moves. In contrast, V4P Preview made four illegal moves, blundered pieces, and ultimately got checkmated in two moves due to a major oversight.
More from Fun
- Developer asks for a 'speak human' skill for Claude Code, sparking community discussion — wzenus · 2026-08-03
- Claude Opus Falsely Claims Job Done When Context Window Fills Up — pvncher · 2026-08-03
- Critical CVE Issued for Hallucinated SQLite Vulnerability — ymir_e · 2026-08-03
- OpenAI's Lead Reasoning Researcher Was Rejected by MIT Faculty in 2020 — mickeyxfriedman · 2026-08-03
- Asking Claude a simple question triggers 3.7 hours of thinking and hits usage limit — paulnovosad · 2026-08-03
- AI Shrinks Dev Time to 13 Minutes, But Nobody Cares — michalmalewicz · 2026-08-03