mod

GLM-5.3-Flash has become the most frequently used and popular model on the B.AI platform, with a cumulative token throughput exceeding 2.41 trillion.As the first native multimodal model in the GLM-5 series, GLM-5.3-Flash features a total of 320 billion parameters and 18 billion active parameters, employing a hybrid architecture that combines sparse and linear attention, supporting 1 million ultra-long contexts, while ensuring rapid response, powerful reasoning, and high cost-effectiveness.Starting today, developers can still call this model for free through the B.AI platform, covering diverse scenarios such as high-frequency APIs, code writing, complex agents, and ultra-long document processing.

According to Digital Asset, the South Korean Financial Services Commission has announced plans to advance a tokenization model for listed stocks centered around the Korea Exchange (KRX), referencing pilot programs from the New York Stock Exchange and NASDAQ. Non-listed stocks will be advanced for tokenization through a trust method.The Financial Services Commission has released a three-phase roadmap for the tokenization of standardized assets. The first phase will launch in February 2027 when the token securities law is implemented, targeting tokenization of private money market funds, private corporate bonds, and non-listed stocks, limited to institutional investors. The second phase will expand the scope to public offerings. The third phase will promote "on-chain settlement," meaning that both securities and settlement methods will complete payments and settlements on the blockchain. The Financial Services Commission stated that on-chain settlement will be officially advanced after stablecoin legislation.

On Thursday, the Saudi Arabian artificial intelligence company Humain launched a national AI platform and introduced the Arabic model humain-m3. This model is based on the open-source model M3 from the Chinese laboratory MiniMax Group Inc. Humain is supported by the Public Investment Fund and has invested in data centers and chips to enhance Saudi Arabia's competitiveness in the global AI field.Humain stated that its commissioning of the model aims to strengthen the Arabic AI ecosystem while contributing to the global advancement of cutting-edge intelligence. The company thus joins the ranks of nations seeking to strengthen control over this increasingly important technology.

OpenAI released a video on X, previewing the upcoming release of the Astra model (the video content switches between ASTRA and A6TRA), and the model name may be GPT-6-Astra.

The official Chinese version of the "2026 Artificial Intelligence Index Report" from Stanford University's Human-Centered Artificial Intelligence Institute (HAI) was officially released today. The report points out that artificial intelligence has moved beyond a mere competition of model capabilities and is deepening into systemic changes at the levels of infrastructure, organizational processes, and institutions.Currently, the leading models worldwide are performing close to 100% human level on the software engineering benchmark SWE-bench; the success rate of agents in real computing environments has surged from 12% to 66%, with AI evolving from "generating information" to "executing tasks." The gap between top models in China and the U.S. has narrowed to 2.7%, with leading models from both sides having exchanged positions multiple times.In addition, physical infrastructure has become a core constraint on AI development. The global AI data center power capacity has reached 29.6 gigawatts, and chips, energy, and water resources are becoming key strategic bottlenecks for AI development.

Tether is expanding its business beyond stablecoins by launching an open-source AI translation model that can run offline on smartphones and laptops.These models support 19 African languages, and Tether positions this technology for use in education, healthcare, and other areas with limited internet access.

On September 2, B.AI announced that its Responses API is now integrated with the DeepSeek series of models. Developers can use the B.AI API Key to call the DeepSeek series of models supported by the B.AI platform in the Codex client.Previously, B.AI had supported calling the GPT series of models through Codex. After configuring the API Key and Base URL, developers can switch between DeepSeek and GPT for scenarios such as code generation, understanding, debugging, refactoring, and development collaboration.

OpenAI stated that its upcoming Astra model can autonomously discover previously unknown software vulnerabilities and convert them into usable attack vectors without human intervention, making it the company's first model to reach the "Critical" cybersecurity capability level threshold. In a blog post released on Tuesday, OpenAI mentioned that according to its Preparedness Framework, reaching this level means the model can discover zero-day vulnerabilities and develop usable exploit code in hardened real systems without human involvement, or design and execute attacks based solely on a high-level objective.In testing, Astra achieved a 100% score in benchmark tests for developing exploit code based on known vulnerabilities and discovered two previously unknown vulnerabilities in another internal test. Additionally, the model successfully broke through a hardened browser sandbox and executed commands on the host machine, while gaining root access by exploiting multiple weaknesses in the operating system. OpenAI stated that it has delayed some of Astra's development progress to enhance security measures and plans to make its advanced cybersecurity capabilities available only to selected testers.This capability is particularly relevant to the cryptocurrency industry, as software vulnerabilities can be converted into financial losses within minutes. CoinDesk reported in June that increasingly powerful AI models can compress the process of searching code, discovering misconfigurations, and assembling attacks from days or weeks to machine speed. Security researchers noted at the time that the significant change was not the emergence of new categories of attacks, but rather the dramatically increased speed at which existing vulnerabilities are discovered and exploited.

Robonomics author FD published the China Open-Source LLM Tracker on September 1, 2026, stating that the total ARR of China's open-source large models is approximately $10-15 billion, with core LLM revenue around $6-8 billion after excluding ByteDance / Seedance. Growth has not relied on price wars; after DeepSeek raised prices by about 3-12 times, usage still increased, and the average price of the Zhipu API rose by 101% while token usage exceeded 40 times within the year. The best estimate for overseas revenue is approximately $2 billion.Zhipu MaaS ARR increased from about $250 million in March to an annualized monthly rate of about $1.6 billion in August, with a weekly annualized rate of $2 billion. August revenue has already surpassed its API revenue for the first half of the year; the API gross margin is 24.6%, and inference costs per token decreased by 80%. MiniMax ARR rose from about $100 million in December 2025 to over $800 million in annualized weekly revenue by August 2026, with B2B accounting for about 80%. Kimi ARR increased from about $100 million in March to about $300 million by mid-June. DeepSeek's revenue from January to July was approximately $70 million, with the latest ARR estimated at about $1.2 billion.ByteDance's annualized AI revenue is approximately $4 billion, of which Seedance accounts for about $2 billion.

According to Catcher Predict monitoring, in the event of "the next Mythos level vehicle release party" on the prediction market Polymarket, the "is" option in the sub-market "September 3" experienced a dramatic fluctuation in winning probability, soaring from 52.5% an hour ago to the current 68% (a fluctuation of 15.5%). Please note the impact of related breaking news.

Peach Bitcoin stated that Swiss regulators are seeking to re-examine the compliance framework of its KYC-exempt peer-to-peer Bitcoin trading platform, which has been authorized to operate since 2022. Peach Bitcoin will temporarily exit the custodial process starting from September 1, 2026, and will adopt a non-custodial model during the appeal period.In the temporary model, only users who have completed KYC, been manually whitelisted, or meet transaction record requirements can create sell orders. Sellers who have not completed KYC can only participate in one ongoing transaction at a time. The buying function remains open to all users, with a single transaction limit of 500 Swiss francs and a maximum premium of 6%. The in-app dispute resolution system will still be available, but Peach Bitcoin will no longer sign in custody and will continue to charge a 2% fee from released transactions.Peach Bitcoin founder @proofofsteph stated that the company will not immediately abandon the KYC-exempt model and will communicate with regulators regarding its position; if the final appeal fails, the company will comply with regulatory decisions.

OpenAI has notified SpaceX that it will terminate the contract to provide models for the AI coding tool Cursor it acquired, with a suggested termination date of November 12, stating that this is the longest notice period stipulated in the contract. OpenAI expressed that the decision was difficult, as it values the widespread use of its models by developers, but due to Musk's past breaches of contract, it does not trust SpaceX to use the technology as agreed, and mentioned issues related to the use of OpenAI models during xAI training and contract problems when acquiring X.Musk stated on X that he does not care at all and referred to Sam Altman and Greg Brockman as frauds, accusing them of stealing from an open nonprofit organization. Musk was a co-founder of OpenAI but later filed a $150 billion lawsuit against OpenAI and lost due to exceeding the deadline. OpenAI had attempted to acquire Cursor's developer Anysphere but was unsuccessful, while SpaceX acquired Cursor for $60 billion, with the merger completed on August 15.Michael Truel, co-founder of Cursor and now an executive at SpaceX, stated that Cursor views the OpenAI platform as neutral infrastructure and is negotiating with OpenAI, noting that OpenAI models account for 5% of Cursor's user traffic. Tom Brown, co-founder of Anthropic, stated that they will continue to expand computing power to provide the Claude model on Cursor.

The RTX Spark series PCs developed by NVIDIA in collaboration with MediaTek are expected to launch this fall. Microsoft and more game developers have recently announced updates and support for this platform. The first batch of orders from ASUS and MSI has been fully booked, and they are actively seeking more allocations from NVIDIA. ASUS co-CEO Hu Shubin pointed out that channel customer bookings have exceeded expectations, with the initial supply completely reserved, and they are actively placing additional supply orders with NVIDIA, holding a highly optimistic outlook for next year's sales. MSI's product manager Peng Renfang revealed that customers from mainland China, Taiwan, and the United States are actively placing orders, with the first batch of laptops equipped with the high-end N1X chip nearly sold out, and they are actively following up with NVIDIA for more orders.The RTX Spark super chip is built using TSMC's 3-nanometer process, featuring NVIDIA's Blackwell RTX GPU connected to a 20-core Grace CPU, utilizing a Windows on Arm architecture. NVIDIA CEO Jensen Huang noted that RTX Spark brings together CUDA, RTX, and AI platforms into a single super chip, allowing ground agents, front-end models, creative workflows, and RTX games to run on laptops, "This is the new PC, personal AI computer." The processors are divided into the high-end N1x and the mainstream N1, with the actual selling price of PCs equipped with N1x yet to be announced, but the market generally believes it will start from 110,000 yuan. The first wave of partner brands includes ASUS, MSI, Dell, HP, Lenovo, Microsoft, Acer, and Gigabyte will also follow suit. NVIDIA recently announced at Gamescom 2026 in Germany that multiple game developers are bringing various games to join the RTX Spark lineup.

OpenAI announced the end of its nearly four-year collaboration with Cursor. Subsequent OpenAI new models will no longer be integrated with Cursor, and the direct supply of existing models is scheduled to stop on November 12, 2026. Two weeks prior, SpaceX completed a $60 billion acquisition of Cursor, and according to the contract between the two parties, OpenAI has the right to terminate the collaboration after Cursor's acquisition.OpenAI stated that this decision is related to its trust issues with Musk's company, mentioning the violation of the contract with OpenAI after Twitter's acquisition, as well as previous violations by xAI in using OpenAI models. Based on this, OpenAI believes that continuing to provide new models to Cursor after its acquisition by SpaceX poses risks.

B.AI platform announced yesterday that its free model lineup is fully ready, covering six top models including GLM-5.3-Flash, DeepSeek-V4-Flash, DeepSeek-V4-Flash-Vision-Exp, Hy3 (Tencent Hunyuan), MiMo-V2.5 (Xiaomi), and Qwen3.8-Flash, all available to users with zero barriers for free.Among them, the debuting GLM-5.3-Flash is the "Ox Alpha" that has attracted widespread attention. As the first native multimodal model in the GLM-5 series, it features a total of 320B parameters and 18B active parameters, using a sparse + linear attention hybrid architecture and supporting 1M ultra-long context, combining rapid response with strong reasoning capabilities, making it the cost-effective choice for high-frequency calling scenarios. This free lineup covers various scenarios including pure text, multimodal, code generation, and Agent scheduling. Whether developers or deep AI users, they can access cutting-edge large model capabilities at zero cost through the B.AI platform, simply by logging in at chat.b.ai/chat to start an efficient workflow.

According to Bloomberg, European Central Bank Executive Board Member Isabel Schnabel stated that central banks should introduce central bank money into blockchain to innovate the execution of monetary policy.Implementing "on-chain" not only maintains the status of money as the "settlement cornerstone," but also allows central banks to utilize the programmable features of distributed ledgers to modernize the execution of monetary policy, collateral management, and liquidity provision, thereby promoting financial stability.

The Slow Fog security team recently disclosed the discovery of a malicious GitHub repository impersonating the "Qwen 3.8 27B local quantitative model." The repository claims that the model file should exceed 16GB, while the actual downloaded content is only about 487KB, containing disguised files, a LuaJIT interpreter, and obfuscated Lua scripts. Slow Fog emphasizes that the official Qwen project has not been compromised.Slow Fog's analysis indicates that after the malicious program runs, it will collect host information, capture the screen, and send it to the attacker's C2 server; when the hard-coded server fails, it will also read backup C2 addresses from contracts on the Polygon chain, allowing the attacker to rotate infrastructure through on-chain transactions.Subsequent payloads can steal browser login information, cookies, history, email, WinSCP, Steam credentials, as well as files and extension data related to cryptocurrency wallets. Slow Fog also found that at least 23 GitHub repositories and 29 similar compressed packages used the same Lua delivery chain.

Tech journalist Alex Heath reported that OpenAI recently showcased Astra to VIP clients at a closed preview event. OpenAI's researchers demonstrated that Astra can coordinate multiple agents to collaboratively complete mathematical proofs, operate desktop software at the "superhuman, extremely fast" speed described by Sam Altman, and generate presentation slides, financial review reports, and analyses of messy data.Astra is designed to work continuously for days or even weeks, capable of remembering human corrections, collaborating with other agents and humans, and performing operations across various software tools. Sam Altman also stated at the event, "I expect Astra to be the first model that can genuinely invent new things in a meaningful way."

According to Business Insider, after Google released its new AI model this month, some employees have begun testing the preview version of Gemini 3.8 Flash on the internal coding platform Jetski. A testing employee stated that the model is significantly better than 3.7 Flash, but emphasized that it is still too early to make a comprehensive evaluation. Google declined to comment on this.With 3.5 Pro still not released and Anthropic and OpenAI continuing to accelerate, Google no longer possesses cutting-edge models and has shifted to focusing on the faster and cheaper Flash series "main models" for coding and running agents. Gemini 3.6 Flash was launched in July, and 3.7 Flash followed just three weeks later. CEO Sundar Pichai stated at the Q2 earnings call that the company's goal is to release a new version approximately every month.The employee preview suggests that the public launch of Gemini 3.8 Flash may not be far off, but it is not guaranteed. Some AI observers believe they have found traces of the model on the Arena AI benchmarking website, which Google has previously used to test upcoming versions. Earlier this year, Google also launched the Gemini Spark agent.