Gemma 4 Multilingual Capabilities Questioned
LChoshen · x · 2026-07-13
A post questions whether Gemma 4 is truly a multilingual model. The author argues that a genuine multilingual LLM should share language-agnostic capabilities, like spatial understanding, across different languages, which Gemma 4 fails to do.
The post mentions having LLMs play 2D board games against themselves in different languages. They found that 2D performance varies depending on the language, indicating that the model's capabilities are not entirely language-independent.
More from Models
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- giffmana: the env being used in training is part of the point — giffmana · 2026-09-11
- awesome-llm-leaderboards: an open-source directory of LLM leaderboards, pricing tables, comparison tools — Last_Establishment_1 · 2026-09-11
- Anthropic claims it works to keep eval environments unidentifiable to models — MaxKannen · 2026-09-11
- Nex N2.5 Pro released on Hugging Face with 407GB of weights — jinnyjuice · 2026-09-11
- RoMa v2 image matching model unveiled in the usual black poster — ducha_aiki · 2026-09-11