Ladies and gentlemen of the NL community, here is the web fallback edition: June 22, 2026, AI Large Model Industry Daily Brief.
Note: This briefing is organized based on web searches and scraped article content. We have distinguished between “June 22 events” and “ongoing events before and after June 22.” Reason: Monday is typically the first trading/release window after weekend information converges, which makes it easy for old news to mix in.
June 22, 2026, AI Large Model Industry Daily Brief
One-Sentence Takeaway
June 22’s theme in the AI large model world is ‘capital revaluation + strategic reflection’: Hong Kong-listed AI concept stocks soared, Microsoft CEO Satya Nadella issued a rare warning about AI monopoly and called for a shift from bleeding-edge competition to inclusive development, WeChat’s “Xiaowei” AI assistant launched in gray mode, and Qi An Xin kicked off its AI Security New Product Week, declaring the dawn of the era of AI agent confrontation.
Comparison with 6.21
| Comparison Module | 6.21 Situation | 6.22 Situation |
|---|---|---|
| Capital Market | GeekPark released DeepSeek financing landscape analysis | HK AI concept stocks soared: MINIMAX-W up 23.9%, Zhipu up 15.1%, Mininglamp up 14.3%, HSI closed down but AI sector bucked the trend |
| GLM-5.2 | “Open source takes the crown” narrative, MIT license, Code Arena #1 | Zhipu stock surged 15.1% in HK; CEO optimistic about China models catching up; no new GLM-5.2 updates |
| DeepSeek Financing | GeekPark confirms ~50B CNY financing and general settlement | No new financing; DeepSeek-V4-Flash #1 globally in API calls for five consecutive weeks |
| Microsoft Stance | Not covered | Nadella publicly warns of AI monopoly: criticizes a few giants “hogging all global learning,” calls for AI to be fairly distributed—may signal a Microsoft AI strategy shift |
| Claude Fable 5 | Unblocking rumors unverified | Nobel laureate John Jumper left DeepMind to join Anthropic |
| New Model/Product Launch | None | WeChat AI assistant “Xiaowei” released in gray mode, main model is in-house WeLM, also calls DeepSeek |
| Security/Governance | None | Qi An Xin kicks off AI Security New Product Week with five new releases to counter AI agent attacks |
| Global Call Volume Data | Not present | Global API calls up for the ninth straight week; China models sweep top four |
1. Hong Kong AI Concept Stocks Surge: Capital Market Starts “Voting”
Core Info
On June 22, HK stocks overall trended lower (HSI down 0.65%, Hang Seng Tech down 1.19%), but AI and large model related industry chains bucked the trend and soared.
Key Details
From Tencent News, June 22, 17:37 article (news.qq.com/rain/a/20260622A0ATDH00):
| Stock | Code | Change |
|---|---|---|
| MINIMAX-W | 00100.HK | Up 23.9% |
| Mininglamp-W | 02718.HK | Up 14.3% |
| Zhipu | 02513.HK | Up 15.1% |
| Enflame (semis) | — | Up 15.3%, record high |
| Jiashin Intl Res. | — | Up 22.4% (tungsten strategic minerals) |
| Xiaomi-W | — | Down 3.5% |
| Kuaishou-W | — | Down 3.6% |
Assessment
Credibility: High
Basis: Complete content from Tencent News, based on HKEX data. This is direct market pricing for domestic large model stocks.
2. Microsoft CEO Nadella Warns of AI Monopoly: Calls for “Ending the Arms Race, Embracing Inclusion”
Core Info
Microsoft CEO Satya Nadella openly criticized the trend of AI industry monopoly in an interview with The Wall Street Journal. This marks his first clear warning against AI concentration as one of the world’s largest AI investors.
Key Details
From AIProductHub summary of The Verge, WSJ, and CNBC coverage (aiproducthub.cn/newsflash/microsoft-satya-nadella-ai-monopoly-warning-wsj-interview):
- Nadella’s original words: “You can’t say, hey, all white-collar jobs will disappear, this could even become weaponized, while we pour all the power into building data centers.”
- Core viewpoint: AI should not become just a tool for a few giants; it should benefit the broader economy and society.
- Context: Microsoft is OpenAI’s largest backer (over $13B invested). There are reports it is considering open models like DeepSeek to cut costs.
- Analyst view: Could signal a major Microsoft AI strategy shift—from betting solely on OpenAI to a more diversified approach.
Assessment
Credibility: High
Basis: Sourced from original reports by The Verge, WSJ, CNBC. This is June 22’s top strategic signal for the AI large model sector.
3. WeChat AI Assistant “Xiaowei” Gray Launch: The National Super-App’s AI Entry
Core Info
WeChat’s native AI assistant “Xiaowei” has started limited gray testing—a new “eye” icon appears on the upper left for some users.
Key Details
From Tencent News, June 22 (news.qq.com/rain/a/20260622A0ATDH00):
- Positioning: Native AI assistant being tested by the WeChat team
- Capabilities: Supports text/voice dialog to operate core WeChat functions, launch mini programs, send messages, check Moments, book services etc.
- Main model: Team’s in-house Chinese large language model WeLM
- Some responses: Calls DeepSeek model
Assessment
Credibility: High
Basis: Tencent News article referencing official WeChat support info. This shows a domestic large model deployed in China’s biggest C-end app.
4. Global Large Model API Data: China Models Sweep Top Four
Core Info
According to the latest OpenRouter data (June 15–21 weekly report), global large model API call volume has risen for nine straight weeks, with Chinese models taking the top four spots.
Key Details
From Tencent News, June 22 (news.qq.com/rain/a/20260622A0ATDH00):
| Metric | Data |
|---|---|
| Global weekly total | 46.7 trillion tokens, up 4.7%, ninth week of growth |
| China weekly volume | 18.81 trillion tokens, eight weeks beating US to lead |
| No.1 | DeepSeek-V4-Flash, 4.94 trillion tokens, five weeks as #1 |
| No.2 | Xiaomi MiMo-V2.5, 3.94T |
| No.3 | MiniMax M3, 3.77T |
| No.4 | Tencent Hy3 preview, 3.63T |
| DeepSeek total | Combined models, 8.65T tokens, 18.5% of world, six weeks at #1 |
Assessment
Credibility: High
Basis: Tencent News quoting OpenRouter data. This directly verifies real usage of domestic models.
5. Nobel Laureate John Jumper Leaves DeepMind for Anthropic
Core Info
On June 19 local time, 2024 Nobel laureate in chemistry John Jumper announced he was leaving Google DeepMind for Anthropic.
Key Details
From Tencent News, June 22 (news.qq.com/rain/a/20260622A0ATDH00):
- Jumper confirmed on platform X
- Won 2024 Nobel Prize in Chemistry alongside DeepMind CEO Hassabis
- Expertise: Protein structure prediction (AlphaFold)
Assessment
Credibility: High
Basis: Tencent News citing Jumper’s X post. Though not a June 22 event (took place June 19), it was included in the daily news brief on June 22.
6. Qi An Xin AI Security New Product Week: Age of “AI Agent Confrontation” Begins
Core Info
On June 22, top domestic cybersecurity company Qi An Xin announced a week-long “AI Era, Offense & Defense First” series with five consecutive launches of new AI security products.
Key Details
From Baidu search result digests (CNR, China Business Journal, People’s Financial News, Sina Finance):
- Launch theme: “Security Intelligence” + “Intelligent Security” double track
- First product: “Threat Analysis Digital Expert,” features “dialogue-based judgment,” covers IOC analysis, sample attribution, APT analysis, etc.
- Qi An Xin chairman Qi Xiangdong: Anthropic’s two new models show that “top models still have inherent security risks”
- Industry view: Cybersecurity has entered the “AI agent confrontation” era
Assessment
Credibility: High
Basis: Consistently summarized by multiple authoritative sources. This is a key signal for the AI security race track.
7. Chance AI: An Input-Free AI App Goes Viral in North America
Core Info
Jiemian News reported that an app called Chance AI is going viral among young people in North America. The report was published June 22.
Key Details
From Sina Finance’s reprint and Jiemian News (finance.sina.com.cn/jjxw/2026-06-22/doc-iniehscs2484639.shtml):
- Product form: No input box, home screen only has a camera button
- Function: After taking a photo, the system analyzes the scene, understands intent, dispatches Agent to explain visual content
- Performance: No.1 worldwide on MMMU visual understanding benchmark (86.07%, outperforms humans at 85.4%)
- Founding team: Shenzhen R&D, Singapore HQ
- Funding: Meitu led multimillion-dollar round; Alibaba Group followed
- Users: Over 200k in 35+ countries, 49.2% next-month retention
- Positioning: “Native AI visual explainer”
Assessment
Credibility: High
Basis: Complete Jiemian News source. Not a core large model release, but demonstrates the move toward “visual gateway” AI apps.
Brief Conclusion
June 22’s core signals are a three-fold mix of ‘capital voting + strategic reflection + application integration’.
The stock surge indicates the market is finally pricing domestic large models; Nadella’s public reflection hints that the world’s biggest AI investor may be shifting strategy; WeChat “Xiaowei” entering gray release means domestic models now have an onramp into China’s biggest C-end app.
Meanwhile, global data proves China models are no longer academic hype—their real token usage now leads the world.
Ladies and gentlemen of the NL community, that concludes the June 22, 2026, AI Large Model Industry Daily Brief. Thank you for reading.
