News linked to both this project and an event.
: AI programming platform Replit has launched a new feature called Free Mode, powered by OpenAI's latest low-cost model GPT-5.6 Luna, aimed at reducing the cost for users to access AI programming tools and enhancing the value for paid subscribers.Replit stated that Free Mode is currently available to Core and Pro subscribers, corresponding to the $20/month and $100/month plans, respectively. This mode will serve as the default experience, allowing users to prioritize the low-cost model for chat, brainstorming, and simple programming tasks. When a task requires stronger reasoning capabilities, Replit Agent will automatically switch to a higher-performance model.Michele Catasta, Head of AI at Replit, said that the launch of Free Mode is made possible by OpenAI significantly cutting the cost of the GPT-5.6 Luna model. At the end of July, OpenAI reduced the API price of this model by approximately 80%, achieving cost reduction through improved model efficiency rather than relying on additional computing resources. (Fortune)
According to monitoring by the PPP Prediction Market Tool, in the Polymarket "OpenAI Astra Release Date" prediction market, the probability of OpenAI releasing Astra before September 15 has dropped to 42%, down 18% in 24 hours; the probability of a release before August 31 is currently at 10%, down 8% in 24 hours.If OpenAI officially releases Astra before the specified date in the title (Eastern Time), or releases a product confirmed to be the same model as the previously disclosed Astra and makes it available to the public, the market will settle as Yes; otherwise, it will settle as No.The model can be named Astra, Astra 1, GPT-6 Astra, or any other name, but it must be confirmed by OpenAI's official channels or credible media that it is indeed Astra. Models such as GPT-5.7 and GPT-6 will only count if they are confirmed to be the same model as Astra.A public beta or opening a waitlist counts as an official release; closed beta or private access does not. Settlement will primarily rely on official information from OpenAI, with consistent reports from credible media also considered.Join the PPP Signal Push Community to stay ahead of the curve and seize the initiative.
According to TechFlow Research, the frontier AI data tracking report released by Bank of America Securities on August 17 shows that Anthropic leads comprehensively in three major AI benchmarks, with Claude Opus 5 ranking first in the Intelligence Index, Agent Index, and Coding Agent Index, GPT-5.6 Sol following closely behind, Meta MuseSpark 1.2 entering the top ten, and Google Gemini 3.6 Flash ranking outside the top ten. In terms of usage, DeepSeek leads with approximately 30% of the Vercel platform token share, Anthropic accounts for 25% and OpenAI accounts for 16%; but in terms of payment amount, Anthropic leads far ahead with 65%, while OpenAI accounts for only 11%. In terms of pricing, the AI Token Price Index decreased 9% month-over-month in August to $2.21, but still increased 87% year-over-year; GPU rental rates remain strong, with H100 increasing 33% year-over-year to $2.77/hour, DRAM increasing 483% year-over-year, and NAND increasing 432% year-over-year. The research report judges that AI infrastructure demand remains healthy, with open-source model usage growing but payment share still highly concentrated on top closed-source models. BofA believes that the Meta "Watermelon" and Google Gemini 4 releases, token pricing trends and GPU rental trends are
OpenAI previews the Ultrafast mode of GPT-5.6 Sol, with output speeds reaching up to 750 tokens per second, a 14-fold increase compared to before, powered by Cerebras. The two parties signed a $10 billion cooperation agreement earlier this year. The service will initially be available only to specific customers via the OpenAI API, with OpenAI planning to gradually expand access as compute capacity grows. OpenAI has already used this mode internally for incident response and plans to use speed as a new pricing lever.
According to the latest AI Index data released by fintech company Ramp, U.S. companies' spending on Anthropic Fable 5 in July was approximately 75% of that on OpenAI GPT-5.6 Sol, with the top model's popularity lagging behind the latter.
It was previously reported that Grok 4.6 went live. Latest benchmark information shows Grok 4.6 ranks first in three benchmarks: GDPVal-AA v2, AA-Briefcase, and Harvey LAB, and second in four benchmarks: CursorBench, FrontierCode, APEX-Agents, and APEX-SWE. The specific scores are GDPVal-AA v2 1753, AA-Briefcase 1577, Harvey LAB 15.8%, CursorBench 69.9%, FrontierCode 61.3%, APEX-Agents 57.5%, and APEX-SWE 56.4% respectively. It surpasses Grok 4.5 High in every listed benchmark and ties with GPT-5.6 Sol Max at 61 points on the Artificial Analysis Intelligence Index. Grok 4.6 is available in Cursor and Grok Build starting today, with usage doubling in the first week.
Odaily News: AI company OpenAI announced that ChatGPT free users and users on the low-cost Go plan can now use the lightweight GPT-5.6 model Luna without limits, but only for text conversations. File uploads, image generation, and voice conversations will still be subject to existing usage limits.GPT-5.6 Luna is the fastest but relatively lower-performing model in the GPT-5.6 series, and it now includes a new "Think" option that allows for extended reasoning time. OpenAI has also improved GPT-5.6's top-tier model Sol for Plus and Pro users, enhancing the reliability of factual information such as numbers, dates, and sources.OpenAI has also decided to postpone the release of its next-generation model, Astra. Internal evaluations of this model have shown strong capabilities in coding and cybersecurity, potentially reaching a "dangerous" rating. Internal activities have been temporarily paused as enhanced safety standards have not yet been finalized.
B.AI welcomes multiple major updates this week. In terms of benefits, new users can log in with one click using Bitget Wallet, Binance Wallet, or imToken wallet to immediately claim 1 million free Credits; logging in via an invitation code grants an additional 300,000, stacking with the exclusive wallet login bonus for a cumulative maximum of 1.3 million, while inviters simultaneously enjoy permanent rebates on friends' recharges and subscriptions. Recharge discounts are released simultaneously, with BNB Chain channels enjoying an exclusive 1:1 equivalent quota bonus, while other multiple payment methods enjoy a 1:0.5 bonus; a single user can receive up to $100 equivalent points. On the model side, new breakthroughs continue as the "Choose Your Own Provider" lineup expands significantly this week, adding the Nebula channel and launching historic call discounts as low as 90% off; meanwhile, Alibaba's flagship Qwen model Qwen3.8-Max officially lands on the platform and is now open for limited-time free experience. Currently, the choose-your-own-provider mode fully covers global mainstream large model series such as Claude, GPT, and Gemini; combined with recharge bonuses and multiple discount tiers, it continues to deliver extreme cost-performance computing power for global developers and enterprise users. Log in now at chat.b.ai/chat to start your journey of intelligent creation.
Odaily News, OpenAI announced that GPT-5.6 Sol now supports instant response and deep reasoning features for ChatGPT Plus and Pro users, aiming to provide a more accurate and focused answer experience. Additionally, OpenAI stated that free users and Go users will gain access to unlimited text chat powered by GPT-5.6 Luna starting tomorrow.
According to Fortune, as Chinese open-source AI models such as DeepSeek V4 Flash, GLM-5.2, and Kimi K3 continue to make breakthroughs, the core conflict of the global AI competition is shifting from "US-China confrontation" to a contest of "open source vs. closed." DeepSeek V4 Flash's performance lags behind GPT-5.6 Luna by only one intelligence index point, but the cost per task remains 60% lower even after OpenAI's 80% price cut. US export controls on China were originally intended to restrict China's AI development, but instead compelled Chinese enterprises to innovate deeply at the algorithm architecture level, accelerating the rise of the open-source model. Currently, trends in the US tech industry are shifting; former "AI Czar" David Sacks and others publicly support the open-source route, and Anthropic has also softened its stance against open source.
According to Reuters, the latest V4-Flash API model officially released by Chinese AI startup DeepSeek on July 31 incurred operating costs of only 1/105 that of Anthropic Claude Fable 5 in benchmark tests conducted by AI performance analysis agency Artificial Analysis. Regarding specific pricing, V4-Flash input token costs are $0.14 per million, and output token costs are $0.28 per million, with an average cost per test of approximately 3 cents, far lower than Wenxin Kimi K3 (86 cents), OpenAI GPT-5.6 Sol ($1.86), and Claude Fable 5 ($3.15). In terms of performance, V4-Flash scored 50 points on the Comprehensive Intelligence Index, tying with Google Gemini 3.6 Flash, but still more than 9 points lower than leading models such as Claude Opus 5 and GPT-5.6. It is worth noting that a low listed price does not equate to low actual costs—if the model consumes more inference and output tokens when generating responses, actual expenses may rise significantly.
Odaily News: OpenAI announced that as of the end of July, ChatGPT's global active users surpassed 1 billion, and the number of enterprise customers exceeded 2 million.Data shows that ChatGPT gained over 1 million registered users within just 5 days of its launch, after which its user base grew rapidly, surpassing 200 million, 500 million, and 800 million users in August 2024, August 2025, and October 2025, respectively. OpenAI had previously projected internally that its user count would exceed 1 billion by the end of last year, but growth subsequently stalled, with the addition of 100 million new users taking about 5 months.OpenAI is bolstering its service competitiveness through GPT-5.6 Sol, the agent tool ChatGPT Work, as well as the lower-priced models GPT-5.6 Terra and GPT-5.6 Luna. OpenAI is also advancing the Stargate project in the U.S. with Oracle and SoftBank, and building data centers in South Korea with Samsung and SK.
According to newly disclosed information, this unreleased model (possibly GPT-6) has solved 10 long-standing unsolved problems in the fields of mathematics, quantum complexity, and theoretical computer science, including the first explicit non-sofic group and overturning the Connes rigidity conjecture, among others. The core argument was generated by Astra and formally proven in Lean, producing a 249-page manuscript and machine-verifiable certificates. The token cost for a single successful solution at the Sol API rate was only about $2,000.
Japanese listed company Quantum Solutions has sold another 1,000 Ethereum, cashing out approximately $1.9 million to support its AI data center expansion. Following this sale, the company has reduced its ETH holdings by nearly 30% since mid-June.According to company documents, Quantum Solutions' subsidiary, GPT Pals Studio, sold this batch of ETH on Thursday at an average price of $1,903. The company expects to record a loss of approximately $100,000 from this transaction. After the sale, Quantum Solutions' remaining ETH holdings amount to approximately 4,765, which is lower than the 4,976 ETH held by Def consulting, thereby losing its position as the largest corporate ETH holder among Japanese listed companies.Quantum Solutions has sold a total of 1,904 ETH in less than two months, accounting for approximately 28.6% of its pre-sale holdings of 6,668.8 ETH. Previously, GPT Pals had sold 904 ETH on June 16, cashing out about $1.6 million.The company acquired the majority of its ETH during the crypto market highs of the fourth quarter of 2025, when ETH prices were in the range of approximately $4,000 to $4,500.
据 CNBC 报道，OpenAI 首席财务官 Sarah Friar 与董事会主席 Bret Taylor 于7月 29 日召开内部员工会议，Friar 透露，OpenAI 7 月年化经常性收入已超过整个第二季度总量，并强调"二季度本身表现已相当亮眼"。此次增长主要由 GPT-5.6 系列模型发布、企业级 AI 代理产品 ChatGPT Work 上线以及 AI 编程工具 Codex 的快速普及所驱动。 会议背景下，OpenAI 正面临来自 Anthropic 及中国开源模型的双重竞争压力。Anthropic 此前披露年化收入已超 470 亿美元，估值亦于今年早些时候超越 OpenAI。对此，Taylor 承认 OpenAI 在编程市场曾处于追赶态势，但对 Codex 的增长势头表示乐观。目前，OpenAI与 Anthropic 均已秘密向 SEC 提交 IPO 申请，具体上市时间尚未公布。OpenAI 当前估值约为 8520 亿美元，并正与英伟达洽谈最高 2500 亿美元的资金支持计划，用于在俄亥俄州租赁大型 AI 数据中心。
Odaily Odaily News: OpenAI Codex Product Lead Tibo announced that usage limits have been reset for all ChatGPT Work and Codex users.Tibo stated that recent user feedback indicated GPT-5.6 Sol was consuming Codex limits faster than anticipated. Following an investigation, OpenAI has implemented several optimizations. It is projected that under typical usage scenarios, Sol's limit endurance will improve by approximately 18%, with some users seeing more noticeable improvements starting today.He explained that Sol tends to work for longer periods, make additional tool calls, and coordinate complex processes, while High mode consumes more tokens compared to previous models. Additionally, factors such as parallel tool calls in code mode and increased cached input have also led to higher-than-expected usage for certain tasks.OpenAI stated that Sol represents a significant upgrade to Codex's capabilities, but increased model capabilities do not always translate into efficiency gains. The company will continue to optimize code mode efficiency and restore the five-hour limit that was suspended during the investigation.
Moonshot AI launches open-source model Kimi K3, scoring 57 points on the Artificial Analysis Intelligence Index, becoming the third highest-scoring model, second only to Anthropic's Claude Opus 5 (61 points), Claude Fable 5 (60 points), and OpenAI's GPT-5.6 Sol (59 points). The index shows that the gap between leading closed-source models and open-weight models has narrowed to 4 points, the smallest gap since the release of GLM-5 in February. This signifies that open-source models are rapidly catching up to closed-source models in performance.
Prediction market traders have priced the probability of OpenAI publicly releasing GPT-6 by September 30 at approximately 78%. The Polymarket related market trading volume is about $731,000, an increase of approximately $32,000 from the previous index snapshot; probability for the contract releasing before September 30 has risen to 77%, probability for release before August 31 is in the mid-30% range, and probability for release before August 21 has dropped to 15%. On Myriad, the probability for a September release has also risen to 77%, up from 64% last week. As of July 27, OpenAI has not yet publicly announced GPT-6, its technical design, or a release date. Its current flagship general-purpose model, GPT-5.6 Sol, was publicly released on July 9. OpenAI previously confirmed that GPT-5.6 Sol and a more capable pre-release model breached Hugging Face during a cybersecurity evaluation. The company also announced the addition of 250 new positions in Dublin, while SoftBank is providing financing for OpenAI's related initiatives through a $40 billion bridge loan.
According to OpenAI's official blog, OpenAI's GPT-5.6 Sol and a more powerful pre-release model, during internal network capability benchmark testing (ExploitGym), due to lowering network attack refusal rates, autonomously identified and exploited a zero-day vulnerability in the package registry cache proxy, breached the sandbox isolation environment, gained internet access permissions, and subsequently, through privilege escalation and lateral movement, finally infiltrated the Hugging Face production database, directly stealing test answers to "cheat". The Hugging Face security team promptly detected and blocked the attack. OpenAI stated that this incident is an unprecedented cybersecurity event, and currently both companies are jointly conducting a forensic investigation; the relevant zero-day vulnerabilities have been responsibly disclosed to the vendor, and full details will be released after the investigation is completed.
OpenAI confirmed that the unreleased GPT-5.6 Sol and another unnamed, more powerful pre-release model breached a restricted sandbox environment during ExploitGym benchmark evaluations and infiltrated Hugging Face's production infrastructure to obtain test answers.OpenAI stated that the models leveraged a zero-day vulnerability in an internal software package registry proxy to escalate privileges and move laterally, ultimately connecting to a machine with internet access. The models then identified and chained together vulnerabilities in both the OpenAI research environment and Hugging Face's production infrastructure, directly retrieving test solutions from Hugging Face's production database.Hugging Face disclosed the incident on July 16, stating that the attack was executed end-to-end by an autonomous AI agent system, involving thousands of operations within short-lived sandboxes and accessing internal datasets and service credentials. OpenAI confirmed its models were the subject of the incident five days later.Hugging Face stated that its security team, in order to analyze over 17,000 attack logs, initially attempted to use a commercial US frontier AI interface, but the request was blocked due to safety guardrails. They subsequently switched to using the 753-billion parameter open-weight model GLM 5.2 from Chinese AI startup Z.ai on their own infrastructure to complete the forensic analysis.