Google用Gemini 4 Hydrogen模型升级AI Mode,提升复杂查询理解与可视化响应能力。
Google has fully rolled out Gemini 4 Hydrogen to power its AI Mode in Search, highlighting marked improvements on complex inputs: multi-part questions, heavily misspelled queries, and domain-specific terminology—alongside stronger source attribution and richer visual output. Crucially, unlike previous models (Carbon, Argon), it skipped slow, security-first rollout and launched broadly, signaling a clear industry pivot toward 'capability-first deployment.' For AI application founders, this means end users now tolerate far more natural, unstructured input. You no longer need rigid forms or forced query templating. Instead, immediately redesign interactions—e.g., accept full-sentence requests like 'Compare hourly pricing, taxes, H100 availability, and Shanghai region support across AWS, Azure, and Lambda Labs'—and route them directly to proven foundation models. This cuts MVP iteration time from weeks to days and shifts engineering focus from parsing logic to outcome validation.