Sakana AI released Fugu Max in September, continuing the lab's focus on efficiency and novel training methods rather than simply building ever-larger models. Sakana is known for approaches inspired by combining smaller models and by nature, betting that cleverness can substitute for sheer scale.
The existence of well-funded labs pursuing efficiency, not just size, matters because it keeps pushing the cost of capable AI down along a different path than the giant-model race.
What it means: bigger is not always better in AI, and the most cost-effective tool for a specific business task is often not the largest, most famous model. It is worth testing a smaller or more specialised model on your actual work before assuming you need the biggest one.