AI Compute Radar has published a "Maker board" ranking 15 labs that contribute to its tracked set of 46 open models, using a composite metric called Heat Score. The ranking is determined by the number of models each lab has in the top 10, with ties broken by best placement and summed 30-day downloads.

  • Alibaba Qwen leads with three top-10 slots and the highest volume of tracked models (11), anchored by Qwen3 0.6B at rank #1.
  • Google ranks second with three slots and the highest mean Heat score (66) among labs with more than two models, led by Gemma 4 31B at rank #2.
  • Zhipu AI holds third place with two slots, showing significant momentum with a +7 Heat point average and +19% download growth.
  • Meta and OpenAI each hold one slot, with Llama 3.1 8B and gpt-oss 120B respectively, though their overall Heat scores have declined recently.

The board distinguishes between breadth (volume of models) and depth (Heat score intensity), noting that recent activity varies significantly among major labs.