Patrick Loeber(@patloeber)
what's your early feedback on Gemini 3.8 Flash? where should we improve?
5.0内容质量
TL;DR · AI 摘要
该推文讨论Gemini 3.8 Flash模型的早期反馈，但缺乏技术深度和具体改进方案，信息密度低。
核心要点
- 用户指出Gemini 3.8 Flash的token效率有较大提升空间
- 高思考模式消耗更多资源但性能与中等模式相似
- 模型控制逻辑存在反直觉设计问题
结构提纲
按章节快速跳转。
- §推文背景
Patrick Loeber发起关于Gemini 3.8 Flash的反馈征集。
- ·用户反馈
ChessBench测试显示模型具备基础棋类能力但存在优化空间。
- ›性能问题
Emilio指出模型存在控制逻辑反直觉问题。
- ›资源消耗
Ugam Kamat发现高思考模式资源消耗过高。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Gemini 3.8 Flash反馈
- 用户反馈
- 控制逻辑问题
- token效率问题
- 性能表现
- 高思考模式资源消耗
金句 / Highlights
值得收藏与分享的关键句。
High thinking mode consumes a lot of tokens and time, at similar performance to medium thinking
the one thing ive noticed is that i really likes to invert controls for some reason
that thing can play some chess!
#Gemini#AI模型#Google#技术反馈
打开原文Patrick Loeber on X: "what's your early feedback on Gemini 3.8 Flash? where should we improve?" / X
[](https://x.com/)
Patrick Loeber on X: "what's your early feedback on Gemini 3.8 Flash? where should we improve?"
-  Patrick Loeber @patloeber what's your early feedback on Gemini 3.8 Flash? where should we improve? 12:13 PM · Sep 3, 2026185.6K Views 962 26 1.2K 98
-  ChessBench @chessbench Sep 3 that thing can play some chess! chessbench.ai/timeline  [](https://x.com/chessbench/status/2095495240388149324/photo/1) 4 3 125 15K
-  Emilio @EmilioSchwaiger Sep 3 the one thing ive noticed is that i really likes to invert controls for some reason. Other than that just from what Ive seen on benchmarks that the token efficiency has lots of room to improve 1 2 52 15K
-  Ugam Kamat @ugamkamat Sep 3 High thinking mode consumes a lot of tokens and time, at similar performance to medium thinking, which uses significantly lower tokens. But, otherwise great job on this release. 1 47 12K
Log in or sign up for X
See what’s happening and join the conversation
Continue with phoneContinue with Apple Continue with Google
or