Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-07-17
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
Kimi K3—featuring 2.8 trillion parameters and million-token context—sets a new benchmark for domestic large language models, with multiple evaluations indicating its overall capabilities are approaching those of leading international models. Simultaneously, the 'Big Three' AI chipmakers—Hygon Information, Moore Threads, and VeriSilicon—announced strong earnings growth forecasts and major order wins on the same day, signaling that China's AI compute chips have officially entered a phase of large-scale commercial deployment [7][11].
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Key Insights
**Kimi K3**, with **2.8 trillion parameters** and **million-token context window**, has redefined the benchmark for domestically developed large language models; multiple evaluations show its comprehensive capabilities are nearing those of top-tier international models. Meanwhile, China's **'Big Three' AI chipmakers**—Hygon Information, Moore Threads, and VeriSilicon—simultaneously signaled robust earnings growth and multi-hundred-million-yuan orders, marking the official onset of mass-scale deployment for domestic AI compute chips [7][11].
## 🚀 Key Updates
- **Kimi K3 Launch: 2.8T Parameters, Million-Token Context, Scores Near Top-Tier Models** [7]: The new model outperforms Opus across multiple benchmark tests, trailing only Fable 5 and GPT-5.6 Sol.
- **Kimi K3 Open-Sources Its 2.8T-Parameter Model—China's AI Gap Narrows to Just 6 Months** [10]: Orange AI notes this open-source move significantly compresses the technological generational gap between Chinese models and global frontrunners.
- **Real-World Frontend Generation Test of Kimi K3: Slow but Effective—Residual Issues Persist After Fixes** [3]: Generating a complete frontend page takes ~1 hour—far longer than competitor Fable (20 minutes); fast-forward/rewind functionality remains unstable.
- **Praising Kimi K3's Aesthetic Sensibility: Independent Generation of Multi-Style HTML+CSS** [5]: The model produces structurally clean, fully functional frontend code (HTML + CSS) in diverse visual styles—all from a single prompt.
- **wigolo: Free, Local, Private MCP Service Empowering AI Agents for Web Search & Scraping** [8]: A lightweight open-source tool delivering private, on-premises web research capability—performance on par with commercial MCP solutions.
- **The 'Big Three' AI Chipmakers Signal a Turning Point** [11]: Hygon Information, Moore Threads, and VeriSilicon jointly announced earnings growth forecasts or billion-yuan-level orders—confirming accelerated rollout of China's AI compute infrastructure.
- **Hands-On Review of WPS Lingxi Professional Edition: AI Emerges as the Next-Generation Office Entry Point** [0]: WPS Office's Lingxi Professional leverages long-term memory and project-level document organization to evolve office AI from 'capable' to 'understanding you'.
- **World Artificial Intelligence Conference 2026: Governance, Academia, and Industry Advance in Tandem** [6]: Signing of the WAICO Cooperation Agreement, launch of the inaugural WAICA academic conference, and debut of hard-tech exhibits—including Huawei's Atlas 950 ultra-node—highlighted the event.
## 🔗 Sources
[0] Hands-On Review of WPS Lingxi Professional Edition: AI Emerges as the Next-Generation Office Entry Point — https://www.bestblogs.dev/article/5d03ac9800?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Starting at ¥95,800 and ¥99,800: Leapmotor Unveils Sedan B01 and SUV B10—Driving Down Prices for 800V Platforms and Dual Zero-Gravity Seats — https://www.bestblogs.dev/article/cc8997e572?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] 8.2 kWh per 100 km: Geely Launches Thunderbolt 16-in-1 Smart Electric Drive System—First Deployed on Galaxy TT — https://www.bestblogs.dev/article/d6d617bf87?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Real-World Frontend Generation Test of Kimi K3: Slow but Effective—Residual Issues Persist After Fixes — https://
Kimi K3, with 2.8 trillion parameters and million-token context window, has redefined the benchmark for domestically developed large language models; multiple evaluations show its comprehensive capabilities are nearing those of top-tier international models. Meanwhile, China's 'Big Three' AI chipmakers—Hygon Information, Moore Threads, and VeriSilicon—simultaneously signaled robust earnings growth and multi-hundred-million-yuan orders, marking the official onset of mass-scale deployment for domestic AI compute chips [7][11].
🚀 Key Updates
- Kimi K3 Launch: 2.8T Parameters, Million-Token Context, Scores Near Top-Tier Models [7]: The new model outperforms Opus across multiple benchmark tests, trailing only Fable 5 and GPT-5.6 Sol.
- Kimi K3 Open-Sources Its 2.8T-Parameter Model—China's AI Gap Narrows to Just 6 Months [10]: Orange AI notes this open-source move significantly compresses the technological generational gap between Chinese models and global frontrunners.
- Real-World Frontend Generation Test of Kimi K3: Slow but Effective—Residual Issues Persist After Fixes [3]: Generating a complete frontend page takes ~1 hour—far longer than competitor Fable (20 minutes); fast-forward/rewind functionality remains unstable.
- Praising Kimi K3's Aesthetic Sensibility: Independent Generation of Multi-Style HTML+CSS [5]: The model produces structurally clean, fully functional frontend code (HTML + CSS) in diverse visual styles—all from a single prompt.
- wigolo: Free, Local, Private MCP Service Empowering AI Agents for Web Search & Scraping [8]: A lightweight open-source tool delivering private, on-premises web research capability—performance on par with commercial MCP solutions.
- The 'Big Three' AI Chipmakers Signal a Turning Point [11]: Hygon Information, Moore Threads, and VeriSilicon jointly announced earnings growth forecasts or billion-yuan-level orders—confirming accelerated rollout of China's AI compute infrastructure.
- Hands-On Review of WPS Lingxi Professional Edition: AI Emerges as the Next-Generation Office Entry Point [0]: WPS Office's Lingxi Professional leverages long-term memory and project-level document organization to evolve office AI from 'capable' to 'understanding you'.
- World Artificial Intelligence Conference 2026: Governance, Academia, and Industry Advance in Tandem [6]: Signing of the WAICO Cooperation Agreement, launch of the inaugural WAICA academic conference, and debut of hard-tech exhibits—including Huawei's Atlas 950 ultra-node—highlighted the event.
🔗 Sources
[0] Hands-On Review of WPS Lingxi Professional Edition: AI Emerges as the Next-Generation Office Entry Point — https://www.bestblogs.dev/article/5d03ac9800?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Starting at ¥95,800 and ¥99,800: Leapmotor Unveils Sedan B01 and SUV B10—Driving Down Prices for 800V Platforms and Dual Zero-Gravity Seats — https://www.bestblogs.dev/article/cc8997e572?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] 8.2 kWh per 100 km: Geely Launches Thunderbolt 16-in-1 Smart Electric Drive System—First Deployed on Galaxy TT — https://www.bestblogs.dev/article/d6d617bf87?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Real-World Frontend Generation Test of Kimi K3: Slow but Effective—Residual Issues Persist After Fixes — https://
← Back to Updates