Fable and the end of the free lunch

The release of Fable has prompted a shift in how developers approach coding and model usage, signaling the end of an era of free, high-performance computing. As Moore's Law slowed, developers had to adapt by optimizing their code and implementing parallelization techniques. Before Fable, it was common to rely on the next model release to fix any issues, regardless of the cost.

However, Fable's release brought a new reality. While Fable is undoubtedly powerful, its cost is prohibitive, and alternatives such as GLM 5.2 are now more affordable and still highly capable. Some believe that the decreasing inference prices will eventually lead developers back to using larger models, but this may not be the case.

The gains from cheaper inference will benefit smaller models like K3 and Qwen, making them more viable options for tasks that previously required larger models. Additionally, Fable's access controls, dynamic degradation, and data retention policies have raised concerns among companies and governments, further driving the need for more cost-effective solutions.