Gemini 3.1 has emerged as a highly efficient AI model, offering performance just a few percentage points below that of the more expensive DeepThink, while being significantly cheaper at nearly 1 oom less. This development highlights a trend in the AI industry where rapid model improvements are drastically reducing the costs associated with high-level intelligence tasks. In particular, the Gemini 3.1 Pro has demonstrated impressive results on the ARC-AGI evaluations, achieving a 98% rating for task completion at a cost of only $0.52 per task, which further pushes the efficiency frontier in AI capabilities.
Gemini 3.1 Pro offers improved efficiency on ARC-AGI benchmarks
