Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-10-05
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
Over the past month, frontier large models have undergone intensive iteration, with four models—GPT-6 Astra, GPT-6.1 Sol, Gemini 4 Argon, and Claude Fable 5.1—released in succession within 30 days, reshaping the model selection landscape once again [0]. Meanwhile, the Trump administration has been active on AI governance, not only signing an executive order to standardize the use of the term "super intelligence" but also announcing the establishment of the "Super Intelligence Force" (SIF) to coordinate federal AI affairs [4][9]...
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Core Insights
Over the past month, **frontier large models** have undergone intensive iteration, with **GPT-6 Astra**, **GPT-6.1 Sol**, **Gemini 4 Argon**, and **Claude Fable 5.1** released in succession within 30 days, reshaping the model selection landscape once again [0]. Meanwhile, the **Trump administration** has been active on AI governance, not only signing an executive order to standardize the use of the term **super intelligence** but also announcing the establishment of the **"Super Intelligence Force" (SIF)** to coordinate federal AI affairs [4][9].
## 🚀 Key Updates
- **Side-by-side comparison of four frontier models released** [0]: MarkTechPost summarizes the differences in suitable tasks among four new models including GPT-6 Astra and Gemini 4 Argon.
- **OpenAI Codex launches 28 days of daily updates** [2]: Commits to releasing one improvement clearly useful to Codex users or a full reset every day.
- **Google pauses open-source vulnerability bounty program** [3]: Due to a surge in AI-automated submissions and the vast majority of reports being invalid, OSS VRP is paused starting October 1.
- **Trump establishes "Super Intelligence Force" (SIF)** [9]: The new agency will coordinate connections between the federal government and AI companies, public interest groups, etc., reporting directly to the president.
- **Meta proposes Context Language Models** [20]: Treats real-time context as model-editable files, uses zero-shot, improves accuracy by 11.4% and reduces FLOPs by 21.5%.
- **Grok 4.7 tops AA Cyber Index** [16]: Musk also endorses its coding ability, saying "Grok 4.7 writes great code" [21].
- **Manus 2.0 integrates H3 Max video model** [8]: The post-trained video model is officially integrated into the Manus 2.0 product.
- **GitHarness uses Git-style branching to manage agent memory** [13]: When requirements change, only the changed parts are redone, reducing token usage by 73.6% on a certain coding task.
## 🔗 Sources
[0] Comparison of four frontier models GPT-6 Astra, GPT-6.1 Sol, Gemini 4 Argon, and Claude Fable 5.1: which to choose for which task — https://aihot.news/items/yy8vov5vuc5pesam5qnf8giih
[2] OpenAI Codex daily updates for the next 28 days — https://aihot.news/items/dmtalxqmeu6koot2yf3kal5nu
[3] Google pauses open-source vulnerability bounty program, citing a sharp increase in AI-automated submissions — https://aihot.news/items/h4ljb95hnvsxrnzrhhadnby1c
[4] TechCrunch podcast explains Trump's super intelligence executive order and voluntary safety commitments — https://aihot.news/items/tlozxgvkj6dknli9f5vy86nx3
[8] Manus 2.0 integrates H3 Max video model — https://aihot.news/items/heag383w3swcknap63ojgjh98
[9] Trump announces establishment of "Super Intelligence Force" (SIF) to coordinate federal government AI affairs — https://aihot.news/items/c64bqsfcqxw8nsneyj0zqsj5k
[13] GitHarness paper proposes using Git-style commits and branches to manage agent memory to handle requirement changes — https://aihot.news/items/h018ba4wueq7bu2l43u
Over the past month, frontier large models have undergone intensive iteration, with GPT-6 Astra, GPT-6.1 Sol, Gemini 4 Argon, and Claude Fable 5.1 released in succession within 30 days, reshaping the model selection landscape once again [0]. Meanwhile, the Trump administration has been active on AI governance, not only signing an executive order to standardize the use of the term super intelligence but also announcing the establishment of the "Super Intelligence Force" (SIF) to coordinate federal AI affairs [4][9].
🚀 Key Updates
- Side-by-side comparison of four frontier models released [0]: MarkTechPost summarizes the differences in suitable tasks among four new models including GPT-6 Astra and Gemini 4 Argon.
- OpenAI Codex launches 28 days of daily updates [2]: Commits to releasing one improvement clearly useful to Codex users or a full reset every day.
- Google pauses open-source vulnerability bounty program [3]: Due to a surge in AI-automated submissions and the vast majority of reports being invalid, OSS VRP is paused starting October 1.
- Trump establishes "Super Intelligence Force" (SIF) [9]: The new agency will coordinate connections between the federal government and AI companies, public interest groups, etc., reporting directly to the president.
- Meta proposes Context Language Models [20]: Treats real-time context as model-editable files, uses zero-shot, improves accuracy by 11.4% and reduces FLOPs by 21.5%.
- Grok 4.7 tops AA Cyber Index [16]: Musk also endorses its coding ability, saying "Grok 4.7 writes great code" [21].
- Manus 2.0 integrates H3 Max video model [8]: The post-trained video model is officially integrated into the Manus 2.0 product.
- GitHarness uses Git-style branching to manage agent memory [13]: When requirements change, only the changed parts are redone, reducing token usage by 73.6% on a certain coding task.
🔗 Sources
[0] Comparison of four frontier models GPT-6 Astra, GPT-6.1 Sol, Gemini 4 Argon, and Claude Fable 5.1: which to choose for which task — https://aihot.news/items/yy8vov5vuc5pesam5qnf8giih
[2] OpenAI Codex daily updates for the next 28 days — https://aihot.news/items/dmtalxqmeu6koot2yf3kal5nu
[3] Google pauses open-source vulnerability bounty program, citing a sharp increase in AI-automated submissions — https://aihot.news/items/h4ljb95hnvsxrnzrhhadnby1c
[4] TechCrunch podcast explains Trump's super intelligence executive order and voluntary safety commitments — https://aihot.news/items/tlozxgvkj6dknli9f5vy86nx3
[8] Manus 2.0 integrates H3 Max video model — https://aihot.news/items/heag383w3swcknap63ojgjh98
[9] Trump announces establishment of "Super Intelligence Force" (SIF) to coordinate federal government AI affairs — https://aihot.news/items/c64bqsfcqxw8nsneyj0zqsj5k
[13] GitHarness paper proposes using Git-style commits and branches to manage agent memory to handle requirement changes — https://aihot.news/items/h018ba4wueq7bu2l43u
← Back to Updates