DeepMind Engineer Asks Community for Gemma Model Ideas

Omar Sanseviero, Developer Experience Lead at Google DeepMind, posted a request for public feedback on desired capabilities for upcoming Gemma models. Replies from researchers and developers highlight needs for base models, stronger tool calling, better VLM judgment, frontend taste, and even more openness in the open-weights ecosystem. The exchange centers on making Gemma more competitive and useful amid frontier open models while keeping post-training refinements for Gemini.