Gemma 4 31B reportedly leads its size class in Medmarks medical AI results

SophontAI says it released additional results for Medmarks, its benchmark and leaderboard for LLM medical capabilities, and finds Gemma 4 31B leads among the mid-size open-source models tested. A commentator sharing the announcement argues that these models have not caught up to frontier models in medicine. They say comparisons of Qwen 3.8 27B with Opus 4.8 or 4.6 may hold for coding, but argue it falls well short of Sonnet 4.5 in medicine.