FlashLM v8.3 (6.5M CORTEX) beats v5.2 Transformer baseline — same 2h CPU, same data
After iterating from v6 to v8.3, FlashLM v8.3 outperforms the Transformer baseline on TinyStories generation quality. Both models trained under identical constraints: Hardware: 2 vCPU / 5GB RAM (free-tier cloud CPU) Time budget: 2 hours wall-clock Dat…