Why have 8B-12B models been dropped?

Model release

Heat trend

Collecting trend data

The percentage is based on available heat signal, not comment count or independent people.

I am a Macbook Pro M4 user with the 16GB of unified ram. The best model I have been able to run on LM Studio is Gemma4 12B QAT, this model is 66 days old. After that the next best thing LM studio suggests is Nemotron 3 Nano 4B and Qwen3.5 9B, which both are 147-161 days old. It also suggests Bonsai 27B which is 10 days old but I dont know if I can support it. Lately it seems all the launches are 27B+, why is that? Is it impossible to make a good SOTA 8B-12B model?