浅眠Sui记
Open in Telegram
这里聊AI/NFT，也发点实用资源 👀channel owner @OvO_GG
Show moreThe country is not specifiedThe category is not specified
6 878
Subscribers
-5024 hours
-3307 days
-3 25230 days
Loading data
Similar Channels
No data
Any problems? Please refresh the page or contact our support manager.
Tags Cloud
No data
Any problems? Please refresh the page or contact our support manager.
Incoming and Outgoing Mentions
---
---
---
---
---
---
Attracting Subscribers
October '26Oct '26
October '26
+13
in 1 channels
September '26
+711
in 3 channels
Get PRO
August '26
+975
in 35 channels
Get PRO
July '26
+21 030
in 5 channels
| Date | Subscriber Growth | Mentions | Channels | |
| 20 October | 0 | |||
| 19 October | +1 | |||
| 18 October | +1 | |||
| 17 October | 0 | |||
| 16 October | 0 | |||
| 15 October | 0 | |||
| 14 October | +1 | |||
| 13 October | 0 | |||
| 12 October | 0 | |||
| 11 October | 0 | |||
| 10 October | +1 | |||
| 09 October | +4 | |||
| 08 October | 0 | |||
| 07 October | 0 | |||
| 06 October | 0 | |||
| 05 October | 0 | |||
| 04 October | 0 | |||
| 03 October | 0 | |||
| 02 October | +1 | |||
| 01 October | +4 |
Channel Posts
| 2 | sticker.webp | 88 |
| 3 | 起床 | 90 |
| 4 | 我要睡觉了，明天记得叫我起床 | 78 |
| 5 | 来一起扎堆
https://t.me/zhadui | 81 |
| 6 | OpenAI Dots 未来或将登陆欧盟
OpenAI ChatGPT 负责人 Tibo
Sottiaux 10 月 18 日回应欧盟用户称，对 Dots 暂不覆盖欧盟表示抱歉，并给出两种解决方案：
1. 在其他地区创建的 Dots 进入欧盟后仍可继续使用。
2. Dots 未来将登陆欧盟。
此前 Dots 在欧洲多地无法使用，欧盟用户多次呼吁 OpenAI 开放该功能。
#OpenAI #Dots #欧盟
——浅眠Sui记 | 103 |
| 7 |  No text... | 101 |
| 8 | 我本来就短😵💫萎了都 | 101 |
| 9 | 陪我短一点 | 102 |
| 10 | 这谁能绷住 | 95 |
| 11 |  No text... | 96 |
| 12 | 太正常了 | 94 |
| 13 | 已联系官方标记为假冒 | 110 |
| 14 |  @fenghaogouu 谨防诈骗 | 113 |
| 15 | Google 开源递归智能体框架 Harness²
Google Research 开源了 Harness²，一个递归优化 LLM 智能体的框架。它通过两层递归在不调整模型权重的情况下提升表现，在 WorkBuddy-Bench 等测试中最高提升 11.5 分。
该框架支持并行与顺序递归模式，已集成多款基准测试，采用 Apache 2.0 协议开源。
#Google #Harness2 #AI智能体 #开源
——浅眠Sui记 | 114 |
| 16 | 英伟达优化 Kimi K3 推理速度，同硬件提速 1.68 倍
SemiAnalysis 发文称，英伟达通过一项上游 vLLM 优化，在 2.8 万亿参数的 Kimi K3 模型上实现了单用户每秒 423 个 token 的速度，相同硬件下快了 1.68 倍。
本次优化由英伟达 CentML 团队使用 Cute DSL 编写了解码专用内核，并通过 Megakernel 融合实现。此外，英伟达也正在为 vLLM 开发适配 DeepSeek v4.1 等模型的工作。
#英伟达 #KimiK3 #vLLM #AI
——浅眠Sui记 | 115 |
| 17 | sticker.webp | 115 |
| 18 | 小北套皮我，没有浮木 | 115 |
| 19 | gg魔了 | 114 |
| 20 | ？ | 117 |
