@Curline1222 @Elaina43114880 5.5是新的预训练模型，参数不小，会有那种“直觉感”，速度很快且大部分时候正确率尚可，这个数数字问题我把快速回答关了基本就没错过了，如果打开错误率会更高，怀疑快速回答是直接用5.5 none或者5.3 instant了
中文
weather will
230 posts
@Elaina43114880 似乎是因为5.5，尤其是在网页版，对简单题目非常轻敌，也不是quick answer设置的原因，自从5.4就这样
中文
RolePlay can sometimes make the model smarter
In this example, GPT-5.5 was set to Thinking Effort: Light, and when asked only in English, the result was counting 10 numbers starting from 11 up to 20.
But when a character name is added—in this case using Chinese, because using a name from a frequently used language can better activate deeper RolePlay—you can see that even under a very low-effort thinking mode, almost comparable to no thinking, the trap was still recognized.
Mario Nawfal@MarioNawfal
WOW! @XFreeze exposed the giant gap between GPT and Grok’s logical reasoning: Same prompt, “Count to 10 starting from 11.” ChatGPT 5.5: obediently goes 11, 12, 13… (zero thinking) Grok 4.3: immediately says 11, 10, 9… and explains why counting backwards is the only logical response. This is why Grok’s reasoning is on another level. Most models follow patterns. @grok actually thinks.
English
@Elaina43114880 @OysterM3 使用的5.4？我以为提示词模型已经换5.5了😂。不过我疑惑的是，image v2的思考强度一共有三个档位，就像pro有xhigh plus没有一样，会不会只有pro才有image v2的high，而plus虽然是延伸但是image v2使用的是medium
中文
I used GPT-5.5 Pro again to try generating a cat by writing SVG code. The Thinking Effort setting was Extended. 🤔
This is the current output. It looks similar to the pre-release results, just slightly better. The gap compared to the results from four days ago is quite large. 😿
I’m not sure if this is just my case or a general trend. If you have Pro access, please try ChatGPT-5.5 Pro Extended with the following prompt and share your result in the comments! 👀
Elaina@Elaina43114880
ChatGPT Pro mode “Create a detailed SVG illustration of a cat.” 👈😻 BEFORE | NOW →_→👉 The left seems to be GPT-5.5 Pro, while the right is GPT-5.4 Pro—you can see how big the gap is! 😾 Hoping to see major improvements in 5 hours! 👌
English
@Elaina43114880 @OysterM3 这个是5.5的思考强度吧，我也不知道这个是否对应image v2，因为image v2最高只有high，按照oai的尿性，我在想是不是free对应low，plus是medium，pro是high，不确定
中文
Workspace agents are now available in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. openai.com/business/works…
English
@synthwavedd @pigeon__s I’m hoping that tomorrow brings gpt-5.5 omni. since the release of gpt-5, oai hasn’t launched any new models with the “-o” suffix. I think it’s about time for another major upgrade to GPT’s voice mode.
English
@pigeon__s will not launch alongside 5.5 unfortunately but shouldn't be that much longer
English
Codex just got a lot more powerful.
Computer use, in-app browser, image generation and editing, 90+ new plugins to connect to everything, multi-terminal, SSH into devboxes, thread automations, rich document editing. Learns from experience and proactively suggestions work. And a ton more.
English
@MemoryQ_Mio @xxxx583996 @tsumetaiame514 挺好的，当时我以为以后就是不温不火的状态了，发生这么大事能转型成功继续发展很厉害了，找个时间回去看看。顺便，新年快乐！
中文
@flower80275 @xxxx583996 @tsumetaiame514 嘉然也在努力，现在as在保原直播内容之余也在转手游，贝拉吃终末地商单，乃琳去年12月开始就是鸣潮一姐，嘉然跟在乃琳后面主打更细致更胎教的方向过鸣潮剧情，甚至助力了鸣潮社区对过往剧情细节的开发，后面来的思诺心宜也吃到了流量，总体在往上走（毕竟在乃琳播潮前团队差到考虑毕业了）
中文
感觉墙内为了避免春晚无聊自发搞的什么大学生春晚、ai春晚、V圈春晚 比真的春晚还无聊一万倍。。
中文
@MemoryQ_Mio @xxxx583996 @tsumetaiame514 那可以啊，转型成功挺好，虽然不看但承载了一些回忆还是希望她们能过好🥲贝拉这个太难绷了，只能说知道mujica第一集那种奇怪的似曾相识的感觉怎么来的了😇现在嘉然怎么样呢？没记错之前算她热度最高了
中文
@flower80275 @xxxx583996 @tsumetaiame514 510后as开始找新方向，乃琳转游戏直播，贝拉去做动画拉片的时候直接变成全网最有资格评价母鸡卡的up主——因为她真的有个被紫色短发女人炸掉的女团😂
至于再后来，那就是去年12月鸣潮拿下玩家之声时被游戏直播的乃琳玩到后一路高歌猛进焕发第二春的故事了，三天前连历史最高团播同接都超了好像是
中文