AI achieves historic breakthrough in mathematical proof: GPT-5.5 Pro solves the 80-year-old 'unit distance problem'; Tencent's Hy-MT2 supports 33 languages across three model sizes with 1.25-bit extreme quantization; Alibaba Cloud's MaaS revenue surges 15× in 5 months—Qwen 3.7 Max tops domestic models and ranks top 5 globally.
## 🔍 Key Insights
**AI-powered mathematical proof** achieves a historic breakthrough: GPT-5.5 Pro disproves the 80-year-old “unit distance problem” in plane geometry; **Tencent Hunyuan’s Hy-MT2** translation model family spans three sizes, supports 33 languages, and enables ultra-low-bit 1.25-bit quantization; **Alibaba Cloud’s MaaS revenue surges 15× in just five months**, with Qwen 3.7 Max crowned top-performing domestic model—and ranked #5 globally—by Artificial Analysis [6][1][9][13].
## 🚀 Major Updates
- **Tencent Hunyuan open-sources the Hy-MT2 translation model family**: three model sizes covering 33 languages, with production-ready 1.25-bit quantization support [1]
- **Alibaba Cloud’s MaaS revenue grows 15× in five months**; Qwen 3.7 Max ranks #1 among domestic models and #5 worldwide in the latest Artificial Analysis benchmark [9][13]
- **GPT-5.5 Pro disproves an 80-year-old mathematical conjecture**; GPT-5 Pro solves the 40-year-old NAG convergence problem—marking AI’s first milestone-level breakthrough in formal mathematics [6]
- **Peking University and Alibaba DAMO Academy jointly release China’s first AI-driven map of wind and solar infrastructure**, published in *Nature* [14]
- **PKU & Baidu launch RepoZero**: the first verifiable, repository-level code generation benchmark—evaluating LLMs’ ability to build complete, functional code repositories from scratch [15]
- **Didi contributes five papers to ICML 2026**, spanning LLM agent evaluation, GUI automation with memory systems, and continual offline reinforcement learning [2]
- **SpaceX’s IPO filing reveals strategic pivot**: ~93% of its valuation now tied to AI compute infrastructure—effectively positioning it as an “AI infrastructure company wearing a rocket shell” [7]
- **DeepSeek formally establishes the Harness team**, dedicated to building a full-stack coding agent toolkit targeting Claude Code–level capabilities [24]
## 🔗 Sources
[1] Tencent Hunyuan Open-Sources New Translation Model Hy-MT2—Three Sizes, 33 Languages — https://www.bestblogs.dev/article/836e339f?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Didi’s Multiple Papers Accepted at ICML 2026—Worth Reading! — https://www.bestblogs.dev/article/61c12724?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[6] An 80-Year-Old Math Conjecture—Just Disproven by GPT — https://www.bestblogs.dev/article/08a1a5af?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] SpaceX: An AI Company in Rocket Clothing — https://www.bestblogs.dev/article/19a0da31?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[9] MaaS Revenue Up 15× in 5 Months: How Alibaba Cloud Cracked the Token Business — https://www.bestblogs.dev/article/e180a25a?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[13] Artificial Analysis Rankings Are Out: Qwen 3.7 Tops Domestic Models—and Lands in the Global Top 5 — https://www
AI-powered mathematical proof achieves a historic breakthrough: GPT-5.5 Pro disproves the 80-year-old “unit distance problem” in plane geometry; Tencent Hunyuan’s Hy-MT2 translation model family spans three sizes, supports 33 languages, and enables ultra-low-bit 1.25-bit quantization; Alibaba Cloud’s MaaS revenue surges 15× in just five months, with Qwen 3.7 Max crowned top-performing domestic model—and ranked #5 globally—by Artificial Analysis [6][1][9][13].
🚀 Major Updates
- Tencent Hunyuan open-sources the Hy-MT2 translation model family: three model sizes covering 33 languages, with production-ready 1.25-bit quantization support [1]
- Alibaba Cloud’s MaaS revenue grows 15× in five months; Qwen 3.7 Max ranks #1 among domestic models and #5 worldwide in the latest Artificial Analysis benchmark [9][13]
- GPT-5.5 Pro disproves an 80-year-old mathematical conjecture; GPT-5 Pro solves the 40-year-old NAG convergence problem—marking AI’s first milestone-level breakthrough in formal mathematics [6]
- Peking University and Alibaba DAMO Academy jointly release China’s first AI-driven map of wind and solar infrastructure, published in Nature [14]
- PKU & Baidu launch RepoZero: the first verifiable, repository-level code generation benchmark—evaluating LLMs’ ability to build complete, functional code repositories from scratch [15]
- Didi contributes five papers to ICML 2026, spanning LLM agent evaluation, GUI automation with memory systems, and continual offline reinforcement learning [2]
- SpaceX’s IPO filing reveals strategic pivot: ~93% of its valuation now tied to AI compute infrastructure—effectively positioning it as an “AI infrastructure company wearing a rocket shell” [7]
- DeepSeek formally establishes the Harness team, dedicated to building a full-stack coding agent toolkit targeting Claude Code–level capabilities [24]
🔗 Sources
[1] Tencent Hunyuan Open-Sources New Translation Model Hy-MT2—Three Sizes, 33 Languages — https://www.bestblogs.dev/article/836e339f?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Didi’s Multiple Papers Accepted at ICML 2026—Worth Reading! — https://www.bestblogs.dev/article/61c12724?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[6] An 80-Year-Old Math Conjecture—Just Disproven by GPT — https://www.bestblogs.dev/article/08a1a5af?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] SpaceX: An AI Company in Rocket Clothing — https://www.bestblogs.dev/article/19a0da31?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[9] MaaS Revenue Up 15× in 5 Months: How Alibaba Cloud Cracked the Token Business — https://www.bestblogs.dev/article/e180a25a?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[13] Artificial Analysis Rankings Are Out: Qwen 3.7 Tops Domestic Models—and Lands in the Global Top 5 — https://www