Google推出Gemini 3 Gemini 3 深思为付费用户服务,促进AI对复杂任务的推理。 Google launches Gemini 3 Deep Think for paying users, boosting AI reasoning on complex tasks.

Google为Google AI Ultra用户推出了Gemini 3深思3, 为复杂的数学、科学和逻辑任务提供高级推理能力。 Google has launched Gemini 3 Deep Think for Google AI Ultra subscribers, offering advanced reasoning capabilities for complex math, science, and logic tasks. 该模型使用并行处理来评估多个假设,在人类的最后一次考试中达到41.0%,在ARC-AGI-2中达到45.1%,并设定了新的基准. The model uses parallel processing to evaluate multiple hypotheses, achieving 41.0% on Humanity’s Last Exam and 45.1% on ARC-AGI-2 with code execution—setting new benchmarks. 它以Gemini 2.5 之前的Gemini 2.5 深思小说成功为基础,通过Gemini 应用程序,与Gemini 3 Pro选择深思小说模式,可以进入Gemini 应用程序。 It builds on prior Gemini 2.5 Deep Think successes and is accessible via the Gemini app by selecting Deep Think mode with Gemini 3 Pro. 该功能目前仅限于付费用户,无法在免费层或其他平台上提供。 The feature is currently limited to paying users and not available on free tiers or other platforms.