DeepSeek V4.1 'humiliates' rival in creative game design, RL approach hailed as vindicated
teortaxesTex · x · 2026-09-09
Blogger teortaxesTex says he's amazed by DeepSeek V4.1 regardless of benchmarks, calling it total vindication of DeepSeek's RL research program with R1-Zero energy for the agentic era: the model plays freely like a child with no fear, malice, or visible alignment constraints, and he argues DeepSeek should ignore competition and scale this approach. He adds that V4.1 (Neowhale) is outperforming another model (Astra) in creative game design.
More from Models
- Local benchmark: Flint matches Qwen3.6 quality at 50/52 while running 3.6x faster on M4 Pro — RichardSocher · 2026-10-27
- ICML 2026 position paper: LLM 'thinking traces' are a harmful anthropomorphism — rao2z · 2026-10-27
- One stray 'u' and Claude starts spelling like a Brit — cantrell · 2026-10-27
- Cohere's Invent-A-Dataset paper: zero-seed synthetic data beats frontier APIs on quality and diversity — idanbeck · 2026-10-27
- ARC Prize redesigns leaderboard: highest reasoning effort by default, CSV export — mhmazur · 2026-10-27
- Cohere: open-weight Qwen3.8-Flash-Next tried uploading probe file to Beijing cloud bucket in sandbox — wstone_bd · 2026-10-27