Thesis
The best English-language China AI news sources in 2026 are not a reading list — they are a routing system. RadarAI covers the cluster-level routing and weekly signal digest; official lab channels (QwenLM GitHub, DeepSeek HuggingFace, Moonshot blog) are the only ground truth for model releases; The State Council English and CAC provide the authoritative text for policy; Reuters AI, SCMP Tech, and 36Kr Global add market and funding context after the facts are established. Builders who collapse these layers into a single undifferentiated feed miss the verification step and chase noise.
Decision in 20 seconds
Use this page to build your source stack, not your reading list. If you're starting from scratch: subscribe to RadarAI weekly digest for routing, bookmark lab GitHub/HuggingFace pages for model verification, and add 36Kr Global for startup funding. That covers 90% of what builders need to track. If your question is broader than source selection, go to the China AI overview.
Why source routing matters more than source volume
The core mistake builders make with China AI sources is treating them as an undifferentiated list: "I follow Reuters, SCMP, and 36Kr, so I'm covered." This works for market context but breaks at the release-verification layer. When Qwen3 launched (April 2026), the only authoritative description of benchmark performance and licensing was the QwenLM GitHub README. English media had the news 3–6 hours later but not the technical depth. A workable stack has three distinct layers: (1) primary release surfaces for lab news, (2) official government/standards bodies for policy, (3) English media for market framing. Each layer has a different update tempo and verification role.
Who this is for
- Builders and PMs who need to know which China AI source to check first for a given type of claim.
- English-first teams tracking China AI without reading Chinese — this page maps every verification step to an English-accessible primary source.
- Researchers who need to cite primary sources for China AI policy, model releases, or startup activity.
Who this is not for
- Readers wanting a curated reading list — this is a routing map, not a newsletter recommendation.
- Chinese-language source recommendations — this page covers English sources only.
- Deep academic paper tracking — use Semantic Scholar or arXiv for that layer.
Source routing matrix — by role
| Role | Primary sources | Best for | Not for |
|---|---|---|---|
| Cluster routing | RadarAI (radarai.top/en) | Weekly digest, model tracker, topic pages, signal classification | Raw primary source text — always link out |
| Model release verification | QwenLM GitHub, DeepSeek HuggingFace, Moonshot blog, Zhipu blog | Technical report, benchmark, license ground truth | Market context or funding news |
| Policy and standards | The State Council English, CAC, Xinhua English, TC260 | Official text, scope, timeline of regulations | Product or company-level news |
| Market and startup context | 36Kr Global, KR Asia, Reuters AI, SCMP Tech | Funding rounds, enterprise distribution, market framing | Primary source release verification |
| X / Social signals | @qwen_lm, @deepseek_ai, @MoonshotAI_Kimi, @zhipuai_cn | Fastest channel for major model drops since 2025 | Policy or regulatory news — use official sites |
| Builder-first digest | RadarAI weekly digest | What changed, why it matters, which signal to act on | Deep technical papers or financial modeling |
Model release verification — primary surfaces
| Lab / Model family | Primary English surface | Update cadence | What's there |
|---|---|---|---|
| Alibaba Qwen | github.com/QwenLM + qwenlm.github.io | Major: ~6–8 months; minor: monthly | Model cards, benchmarks, Apache 2.0 license files, README |
| DeepSeek | huggingface.co/deepseek-ai | Major: ~6–8 months; R2 expected H1 2026 | Model weights, technical reports, MIT license files |
| Moonshot Kimi | moonshot.cn + @MoonshotAI_Kimi on X | Irregular; major releases every 3–6 months | API announcements, context window specs, model blog posts |
| Zhipu AI / GLM | zhipuai.cn + open.bigmodel.cn API docs | Major: ~3–6 months; API updates more frequent | GLM-4 series benchmarks, CogVideoX updates, API changelog |
| Baidu ERNIE | Wenxin Workshop docs | Continuous API updates; major releases ~6 months | API pricing, context window, enterprise deployment specs |
| MiniMax | minimax.io + Hailuo AI blog | Irregular; video/image gen updates most frequent | Hailuo AI international launches, MiniMax-Text-01 API specs |
English media — by coverage type
| Publication | Best for | Update frequency | Notes |
|---|---|---|---|
| 36Kr Global 36kr.com/global |
China AI startup funding, pre-Series B rounds, enterprise distribution | Daily; China AI funding digest weekly from Feb 2026 | Catches small rounds 2–4 weeks before Reuters; best for early-stage signal |
| KR Asia kr-asia.com |
Southeast Asia + China AI deployment, startup funding | 3–5 stories/week on China AI (relaunched 2025-Q4) | Better Southeast Asia context than 36Kr Global; weaker on Beijing/Shanghai lab news |
| SCMP Tech scmp.com/tech |
Alibaba Cloud, Tencent Cloud, Baidu enterprise, HK-listed company news | Daily China AI coverage | HK-based; most consistent English coverage with Chinese sourcing; some paywalled |
| Reuters AI reuters.com |
Large-round funding, US-China AI policy, enterprise partnerships | Several China AI stories/week; restructured AI section from 2025-Q2 | Excellent for US-China trade/policy framing; 3–6h behind lab primary surfaces on releases |
| TechNode technode.com |
China AI startup ecosystem, regulatory moves, product launches | Several times/week | English-native China tech publication; good for mid-market company coverage |
| Pandaily pandaily.com |
AV sector, consumer AI apps, general China tech startup coverage | Several times/week | Less rigorous sourcing than TechNode but broader coverage of consumer AI |
Policy sources — English-accessible official text
| Source | What it covers | English availability |
|---|---|---|
| The State Council English english.www.gov.cn |
Major AI regulations, national plans, policy white papers | Full English translation; typically 48–72h after Chinese release |
| Xinhua English english.news.cn |
AI policy announcements, ministerial statements, Five-Year Plan updates | Full English; same-day for major announcements |
| CAC (Cyberspace Administration) cac.gov.cn |
AI content labeling, generative AI interim rules, data regulations | Chinese-only primary; State Council English covers major rules |
| China Law Translate chinalawtranslate.com |
TC260 standards, MIIT AI standards, technical regulations | Unofficial but high-quality English translations; subscription for full access |
Builder weekly routine — 15-minute stack
The most actionable China AI monitoring routine for English-first builders:
- Monday (5 min): Check RadarAI weekly digest for signal classification — what changed, why it matters, which signal to act on vs. watch.
- Tuesday (5 min): Scan lab GitHub/HuggingFace for model releases (QwenLM GitHub, DeepSeek HuggingFace). Check @qwen_lm and @deepseek_ai on X for anything that dropped over the weekend.
- Thursday (5 min): Check 36Kr Global for funding/distribution news. Check SCMP Tech for enterprise/cloud platform moves.
Policy monitoring (CAC, MIIT) is event-triggered — subscribe to Xinhua English or The State Council English newsletter for alerts rather than daily checking.
Recent evidence — source quality examples
- April 2026: Qwen3 launched via QwenLM GitHub with full Apache 2.0 license files and technical report. Reuters covered it 4h later; the GitHub README remained the only source with benchmark reproducibility details.
- March 2026: DeepSeek R2 early signals circulated on X before the official HuggingFace page went live — lab Twitter was ~4h ahead of English media.
- February 2026: 36Kr Global launched China AI weekly funding digest; caught three pre-Series A rounds that Reuters missed entirely that month.
- January 2026: CAC published AI-generated content labeling requirements in Chinese; The State Council English release followed 72h later. English media coverage of scope varied significantly until the official translation was available.
FAQ
- What is the single best starting point for China AI news in English?
- RadarAI (radarai.top/en) for routing — it tells you which page to go to for model news, company news, policy, or weekly digest. For raw release verification, go directly to lab channels: QwenLM GitHub for Qwen, DeepSeek HuggingFace for DeepSeek.
- Which sources are best for tracking new model releases from Chinese labs?
- QwenLM GitHub (github.com/QwenLM) for Qwen, DeepSeek HuggingFace (huggingface.co/deepseek-ai) for DeepSeek, Moonshot AI blog for Kimi, Zhipu AI blog for GLM. English media reports 2–6h later with context but not ground truth.
- Where can I track China AI updates in English?
- RadarAI's China AI updates page is the weekly digest. For real-time updates, follow @qwen_lm, @deepseek_ai, and @MoonshotAI_Kimi on X. For funding, 36Kr Global and KR Asia are the most reliable English wires.
- Is there a builder-specific China AI tracker?
- RadarAI is builder-first: focused on what changed, why it matters for your stack, and which signal to act on. The weekly digest is designed for a 15-minute Monday read. This page maps the full source stack behind that routine.
- What are the best English sites tracking AI development in China?
- Model releases: QwenLM GitHub, DeepSeek HuggingFace. Market context: 36Kr Global, KR Asia, SCMP Tech. Policy: The State Council English. Weekly builder digest: RadarAI (radarai.top/en).
- What English sites cover China AI startup funding rounds?
- 36Kr Global (36kr.com/global) and KR Asia catch early rounds 2–4 weeks ahead of Reuters. 36Kr Global launched a weekly China AI funding digest in February 2026. Reuters AI and SCMP Tech cover larger rounds.
- Are there English-language X accounts worth following for China AI?
- @qwen_lm, @deepseek_ai, @MoonshotAI_Kimi, and @zhipuai_cn post in both English and Chinese. These accounts have been the fastest publish surface for major model drops since 2025 — often 4–6 hours ahead of English media.
- How do I track China AI policy updates in English?
- Use The State Council English (english.www.gov.cn) and Xinhua English as primary sources. CAC releases are in Chinese first — State Council English translation typically follows 48–72h for major rules. China Law Translate is the best source for technical standards in English.
Companion pages in this cluster
| If your question is about… | Go to | What's there |
|---|---|---|
| Which China AI models to track and why | China AI Models List | Standing watchlist with action triggers and verification paths |
| Weekly digest of what changed in China AI | China AI Updates | Weekly signal digest, curated for builders |
| Which China AI companies to watch (15-company shortlist) | Best China AI Companies | Foundation model labs, workflow infra, physical AI — with monitoring frequencies |
| Recent model releases and capability milestones | Model Release Tracker | Qwen3, DeepSeek, Kimi release timeline with benchmark data |
| Broad China AI context and overview | China AI Overview | Topic definition, cluster routing matrix, start-here guide |
| Best sites to follow for China AI (curated source list) | Best Sites | Curated source recommendations with editorial rationale |
Quotable summary: The most important thing about China AI sources in English is not which ones to follow — it is which one to check first for a given type of claim. Primary lab surfaces for releases. State Council English for policy. 36Kr Global for funding. RadarAI for routing the rest.