Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-10-27
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
OpenAI has launched GPT-6.1 Terra at a "rock-bottom price," dramatically lowering the cost of deploying Agents, while its model solved multiple mathematical problems overnight, including the Local Smoothing Conjecture, marking AI's dual-track advance from "high-value experimentation" to "large-scale daily applications" and "frontier scientific research production tools" [1][2]. Meanwhile, Ant Group's Lingguang 2.0 demonstrated the "managing" capabilities of a Chinese-style Personal Agent, while NVIDIA and Meta's $3 billion partnership, Anthro...
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Core Insights
**OpenAI** has launched **GPT-6.1 Terra** at a "rock-bottom price," dramatically lowering the cost of deploying Agents, while its model solved multiple mathematical problems overnight, including the **Local Smoothing Conjecture**, marking AI's dual-track advance from "high-value experimentation" to "large-scale daily applications" and "frontier scientific research production tools" [1][2]. Meanwhile, **Ant Group's Lingguang 2.0** demonstrated the "managing" capabilities of a Chinese-style Personal Agent, while capital moves such as **NVIDIA and Meta**'s $3 billion partnership and **Anthropic**'s sprint toward IPO signal that the infrastructure and business landscape of the Agent era are accelerating into shape [0][10][11].
## 🚀 Key Developments
- **GPT-6.1 Terra launched at rock-bottom price, the working model for the Agent era arrives** [1]: OpenAI's new model halves the price while approaching flagship performance, greatly reducing Agent deployment costs.
- **OpenAI model solves the King's conjecture overnight** [2]: Proved the Local Smoothing Conjecture in one go, overturned a nearly 30-year-old ETH hypothesis, and constructed a rigorous proof for Yang-Mills theory.
- **Ant Group Lingguang 2.0 hands-on test: Chinese-style Muse can "manage things"** [0]: Unlike errand-running Agents, it can proactively follow up on long-horizon dynamic tasks such as pet medical care and Double 11 price comparisons.
- **Hands-on tutorial for connecting 15 national-level apps to Lingguang 2.0** [7]: Covers integration methods including QR code authorization, official MCP, Skill, and CLI, with step-by-step configuration prompts.
- **NVIDIA and Meta expand Muse model collaboration with $3 billion investment** [11]: Optimizing inference and training performance around the Muse series, exploring OpenShell and Sentry monitoring directions.
- **Baidu open-sources Unlimited OCR 2** [13]: 3.34B parameters, MIT license, can parse hundreds of pages of documents in a single request, with decoding overhead essentially not growing with page count.
- **vLLM releases 27B open-source multimodal decision model d3-ultra** [12]: Apache-2.0 license, does not generate text, reads input once and returns probabilities for each question.
- **Qwen launches question finding, photo-based Q&A, and learning profiles** [9]: Supports precise question finding with free downloads, provides solution steps according to grading standards, and automatically files incorrect answers into learning profiles to pinpoint weak points.
## 🔗 Sources
[0] Chinese-style Muse can manage things! I handed pet insurance policies, Double 11 price calculations, and weekend travel all to Lingguang 2.0 — https://www.bestblogs.dev/article/c71179d8c3?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Just now, GPT-6.1 Terra launched at rock-bottom price, the working model for the Agent era arrives — https://www.bestblogs.dev/article/ec30fdfc99?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] The King's conjecture, solved by AI overnight — https://www.bestblogs.dev/article/0a2628d5c9?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] Connecting 15 national-level apps to Ant Group Lingguang 2.0: A hands-on tutorial for Personal Agent — https://aihot.news/items/d3m8z4cncilvrbdfjmf7e0m82
[9] Qwen launches question finding, photo-based Q&A, and learning profile features to help students precisely
OpenAI has launched GPT-6.1 Terra at a "rock-bottom price," dramatically lowering the cost of deploying Agents, while its model solved multiple mathematical problems overnight, including the Local Smoothing Conjecture, marking AI's dual-track advance from "high-value experimentation" to "large-scale daily applications" and "frontier scientific research production tools" [1][2]. Meanwhile, Ant Group's Lingguang 2.0 demonstrated the "managing" capabilities of a Chinese-style Personal Agent, while capital moves such as NVIDIA and Meta's $3 billion partnership and Anthropic's sprint toward IPO signal that the infrastructure and business landscape of the Agent era are accelerating into shape [0][10][11].
🚀 Key Developments
- GPT-6.1 Terra launched at rock-bottom price, the working model for the Agent era arrives [1]: OpenAI's new model halves the price while approaching flagship performance, greatly reducing Agent deployment costs.
- OpenAI model solves the King's conjecture overnight [2]: Proved the Local Smoothing Conjecture in one go, overturned a nearly 30-year-old ETH hypothesis, and constructed a rigorous proof for Yang-Mills theory.
- Ant Group Lingguang 2.0 hands-on test: Chinese-style Muse can "manage things" [0]: Unlike errand-running Agents, it can proactively follow up on long-horizon dynamic tasks such as pet medical care and Double 11 price comparisons.
- Hands-on tutorial for connecting 15 national-level apps to Lingguang 2.0 [7]: Covers integration methods including QR code authorization, official MCP, Skill, and CLI, with step-by-step configuration prompts.
- NVIDIA and Meta expand Muse model collaboration with $3 billion investment [11]: Optimizing inference and training performance around the Muse series, exploring OpenShell and Sentry monitoring directions.
- Baidu open-sources Unlimited OCR 2 [13]: 3.34B parameters, MIT license, can parse hundreds of pages of documents in a single request, with decoding overhead essentially not growing with page count.
- vLLM releases 27B open-source multimodal decision model d3-ultra [12]: Apache-2.0 license, does not generate text, reads input once and returns probabilities for each question.
- Qwen launches question finding, photo-based Q&A, and learning profiles [9]: Supports precise question finding with free downloads, provides solution steps according to grading standards, and automatically files incorrect answers into learning profiles to pinpoint weak points.
🔗 Sources
[0] Chinese-style Muse can manage things! I handed pet insurance policies, Double 11 price calculations, and weekend travel all to Lingguang 2.0 — https://www.bestblogs.dev/article/c71179d8c3?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Just now, GPT-6.1 Terra launched at rock-bottom price, the working model for the Agent era arrives — https://www.bestblogs.dev/article/ec30fdfc99?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] The King's conjecture, solved by AI overnight — https://www.bestblogs.dev/article/0a2628d5c9?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] Connecting 15 national-level apps to Ant Group Lingguang 2.0: A hands-on tutorial for Personal Agent — https://aihot.news/items/d3m8z4cncilvrbdfjmf7e0m82
[9] Qwen launches question finding, photo-based Q&A, and learning profile features to help students precisely
← Back to Updates