Does AI Pro have deteriorated performance compared to Ultra?

,

Gemini 3.1 has deteriorated performance in AI Pro when compared to AI Ultra even when having enough quotas. Does AI Pro use a stricter token budget per prompt?