Qwen 2.5 27B raises questions about scaling laws
Notrx73 · reddit · 2026-08-18
The performance of the Qwen 2.5 27B model sparks debate: 1) Are scaling laws dead? 2) Are parameters in frontier models mostly fluff? 3) How was this achieved—via distillation or RL training efficiency?
More from Models
- Harvard's Zak Kohane finds 5 AI detectors all flag his own writing as AI — zakkohane · 2026-08-18
- Sakana AI releases Japanese-specialized reasoning model Sakana Namazu — hardmaru · 2026-08-18
- Claim: DeepSeek V4 Beats Fable with J-Space Plugin Fixes — jmorant555 · 2026-08-18
- Running a fully local AI podcast station with Qwen — sysadmin420 · 2026-08-18
- Built a game in two prompts with Qwen 3.8 — lordekeen · 2026-08-18
- User claims GigaChat 3 Ultra is the best Russian LLM right now — flowersslop · 2026-08-18