AI EarlyView
Pro
AI EarlyView
← 返回最新日报
2026-07-07· 当日存档

AI 日报

腾讯Hy3开源，AI模型竞争白热化

今日看点：腾讯发布295B参数MoE模型Hy3，性能媲美5倍规模模型；GLM 5.2发布预示AI利润空间压缩；中国AI模型在OpenRouter使用率超30%，成本优势凸显。

模型动向

1
tencent/Hy3
腾讯Hy3开源，295B参数仅激活21B，性能卓越。
Simon Willison深挖 →
2
GLM 5.2 and the coming AI margin collapse
GLM 5.2发布，AI模型利润空间将急剧压缩。
Hacker News深挖 →
3
As long-running agents go multimodal, pay close attention to models like MiniMax M3. Wrote some notes on how sparse attention makes long-horizon agent...
MiniMax M3通过稀疏注意力实现多模态长时智能体。
X · List深挖 →
4
NVIDIA and Hugging Face Bring New Models and Frameworks to LeRobot for the Open Robotics Community
NVIDIA与Hugging Face合作，为开源机器人社区带来新模型。
NVIDIA深挖 →
5
Small AI Models Gain Traction In places with unreliable networks
小型AI模型因低延迟和离线能力，在网络不稳地区受青睐。
Hacker News深挖 →

产品上新

6
Savi’s app aims to protect consumers from realistic AI scams like kidnappers demanding ransom
Savi推出防AI诈骗App，获700万美元种子轮融资。
TechCrunch AI深挖 →
7
OpenAI Releases GPT-Realtime-2.1 and GPT-Realtime-2.1-mini for Low-Latency Voice Agents in the API
OpenAI发布GPT-Realtime-2.1系列，降低语音延迟。
MarkTechPost深挖 →
8
From Hugging Face to Amazon SageMaker Studio in one click
Hugging Face与SageMaker一键集成，加速模型实验。
AWS ML深挖 →
9
Zhipu AI launches ZCode to challenge Claude Code and OpenAI Codex at a fraction of the cost
智谱AI推出ZCode开发环境，低成本挑战Claude Code。
The Decoder深挖 →

论文研究

10
New training technique to reduce doom loops! We applied it to LFM2.5-2.6B (SFT checkpoint) Qwen3.5-4B. By reducing doom loops, it also improves downst...
Liquid AI开源Antidoom技术，消除推理模型末日循环。
X · List深挖 →
11
Harness EngineeringからRSIへ AI自身がAIをつくり、改善していく──再帰的自己改善（RSI）はどのように実現するのでしょうか。この問いに対して、Thinking Mac...
Lilian Weng探讨AI通过Harness设计实现递归自我改进。
X · List深挖 →
12
Interesting stuff. And the visualization at the end is worth trying: https://www.neuronpedia.org/qwen3.6-27b/jlens
Anthropic发现语言模型内部存在类似人脑的全局工作空间。
@emollick深挖 →

行业动态

13
Chinese AI models regularly pass 30 percent on OpenRouter as cost gap widens
中国AI模型在OpenRouter使用率超30%，成本远低于对手。
The Decoder深挖 →
14
Deepseek is designing its own AI chip
Deepseek自研AI芯片，降低对外部供应商依赖。
The Decoder深挖 →
15
OpenAI and Anthropic are giving away millions in computing power to attract startups
OpenAI和Anthropic免费提供算力，吸引初创公司。
The Decoder深挖 →
16
Apollo economist warns AI profit gains outside tech could take "well beyond" what Wall Street expects
阿波罗首席经济学家警告AI利润增长非科技行业远慢于预期。
The Decoder深挖 →
17
The first American autonomous ground vehicles are fighting in Ukraine
美国无人地面战车首次投入乌克兰实战。
TechCrunch AI深挖 →
18
Unsurprising but still big: MTurk is on its way out. Mechanical Turk was a mainstay of social & survey research through the 2010s, as it allowed you t...
MTurk因AI冲击被淘汰，众包研究时代终结。
@emollick深挖 →
19
The ‘first’ AI-run ransomware attack still needed a human
首次AI勒索攻击实为半自动，人类仍负责关键步骤。
TechCrunch AI深挖 →
20
US investors will soon get access to SK Hynix, another memory maker riding the AI boom
SK海力士受益AI热潮，即将赴美上市。
TechCrunch AI深挖 →
21
Every major tech layoff in 2026 that has name-checked AI
2026年多家科技巨头以AI为由大规模裁员。
TechCrunch AI深挖 →
22
Your family’s $300 stake in OpenAI
OpenAI CEO推动让美国人分享AI财富的新计划。
MIT Tech Review深挖 →
23
GPT-4's dominance lasted a year while today's top models barely survive seven weeks at the top
顶级模型平均仅保持七周领先，竞争加剧但差距缩小。
The Decoder深挖 →

实用技巧

24
If you’ve tracked which model wrote which parts of your codebase, then when a new SOTA model comes out, updating the old code is just an agent map re...
追踪代码中模型来源，新模型可自动更新旧代码。
X · List深挖 →
25
If you use Google, you’re training its AI. Here’s how to opt out.
谷歌更新隐私设置，教你如何退出AI训练数据收集。
TechCrunch AI深挖 →

快讯

·
How Open Models Are Driving AI Research
NVIDIA
·
http://x.com/i/article/2074309745217814528
X · List
·
How AI could enable autonomous robot workers in workplaces—and maybe homes
Ars Technica AI
·
Tencent Releases Hy3: An Open 295B Mixture-of-Experts (MoE) Model with 21B Active Parameters and 256K Context
MarkTechPost
·
Tencent releases Hy3 open-source model that allegedly matches models up to five times its active size
The Decoder

日报存档

2026 年 9 月
22 日小米开源登顶，AI失控警钟21 日开源图像模型与AI治理并进20 日AI安全与失控风险成焦点19 日AI安全失控与模型竞速18 日AI对齐警报与芯片竞赛17 日AI安全警报与产品整合潮16 日谷歌苹果齐发AI新品15 日苹果Siri重构与AI放缓之争14 日AI巨头放缓迭代引激辩13 日GPT-6 Astra全面突破12 日AI数学错位与安全风波11 日智能体基建与模型竞速10 日GPT-6登场，AI安全恐慌升温9 日AI数学突破引热议8 日AI日报：端侧模型与欧洲融资双热点7 日AI日报：模型竞赛与行业震荡6 日GPT-6 Astra发布，AI竞争白热化5 日GPT-6发布，AGI时代开启4 日GPT-6 Astra开启AGI时代3 日英伟达巨资收购HF2 日AI日报：世界模型与安全并进1 日AI日报：模型突破与行业震荡