Gemini 3.0 might do well even without web searches. The lesson from gpt 4.5 and Gemini 3 seems to be that scaling model size (even if you use sparse MoE) allows

Gemini 3.0 might do well even without web searches. The lesson from gpt 4.5 and Gemini 3 seems to be that scaling model size (even if you use sparse MoE) allows you to capture more long-tail knowledge. Some of Humanity's Last Exam also seems to be explicitly designed to test this long-tail obscure knowledge extraction, and models have been steadily chipping away at it.