Gemini 4 Hydrogen 成为 Gemini 应用默认模型
谷歌宣布 Gemini 4 Hydrogen 成为 Gemini 应用的新默认模型,并承诺比 Gemini 3.8 Flash 更少遗忘要求、响应更快且更准确。在 TechRadar 作者针对太阳能板规划的单项测试中,该模型记住了此前的对话,提供英国零售商的具体价格,生成交互式节省预测和合规检查清单。作者还称完成工作流所需的提示和纠正更少、速度明显提升,但这并非受控基准测试。
如果其跨对话连续性和交互输出能达到报道中的表现,Gemini 4 Hydrogen 可让 Gemini 应用更有效地处理此前需要反复补充上下文和纠正的多步骤项目。
Google's new Gemini 4 Hydrogen, the new default model in the Gemini app, promises to cut out the forgetfulness usually associated with AI and keep you on task. It also promises to respond faster and more accurately than 3.8 Flash.
One of the most annoying things I’ve found about using AI on the day-to-day, is having to remind it of my specific preferences and rules when I write prompts. We’ve all tried to give AI a set of guidelines in a prompt, such as a word count or specific inclusions, only to find that the final result is completely lacking them. Or sometimes, Gemini Flash would just make up an answer.
As someone who has frequently had to enter a prompting battle to get an acceptable result out of AI, I was curious to put Gemini 4 Hydrogen to the test to see if it could do better — and I was pleasantly surprised.
No more blank slate
To put Gemini 4 Hydrogen to the test against its predecessor, 3.8 Flash, I decided to continue a project I started during the height of the summer heatwave. I’ve long been considering installing some kind of solar panel setup to subsidize my energy bills, and given how much sun the UK actually got this year, and how much money I spent running an air conditioner, it really seems like a no-brainer.