23:01
2026-09-29
dev.to
artificial-intelligence
Darwin-180B-RSI: VIDRAFT's Recursive Self-Improvement MoE Model Claims Five Leaderboard Tops
Korean Pre-AGI startup VIDRAFT released Darwin-180B-RSI, a 180-billion-parameter mixture-of-experts reasoning model built on Qwen3.8-Flash-Next that combines selective model merging with recursive sel…