Pro Writers Still Crush GPT-6

🔥 What happened Vulsar AI benchmarked 24 LLMs against human writers across 475 long-form prompts. GPT-6 Astra hit 87.8% predicted win rate — barely beating amateurs (86.6%) but losing badly to professionals. 💡 Why it matters Smaller models collapse: Gemma 4 scores 10.9%, DeepSeek V4.1 Flash 19.8%. Pros generate 2,592 tokens per prompt — nearly double GPT-6 Astra's 1,537. Coherence over multi-chapter prompts remains a killer. ⚡ Our take Frontier models beating hobbyists isn't news. Losing to pros by a wide margin is. Anyone claiming AI replaces professional writers hasn't read this leaderboard.