Gemma 4 lands with grounded personality

Hot take: Gemma 4 seems to be winning some users over not just on capability, but on conversational temperament, which matters a lot in day-to-day local use.

• –The thread is explicitly about qualitative feel, not benchmark performance, so this is anecdotal but still useful signal.
• –The poster describes the model as “bearable” even without a system prompt, which suggests the base chat tuning is less hostile or brittle than many local models.
• –The strongest praise is that it feels grounded and not overly sycophantic, which is a rare sweet spot for assistant-style models.
• –The same comment also notes it stays assistant-oriented by asking follow-up questions, so it is not just “neutral,” it is actively conversational.
• –This kind of feedback matters because personality is often the difference between a model people test once and one they actually keep using.