openai

OpenAI documentation shows that eligible ChatGPT Pro and Business subscriptions can receive free API credits for building applications and agents within their own API organization. The credits are automatically issued based on the subscription, and any unused portion expires at the end of the billing cycle without carrying over; when there are valid free credits, there is no need to bind an API payment method. In terms of monthly credits, Pro's tiers of $100, $200, and $500 correspond to $100, $200, and $500 respectively; Business Standard provides $20 per paid seat, and Business Premium provides $100 per paid seat.The initial issuance is calculated proportionally based on the remaining time in the current subscription billing cycle and expires at the end of that cycle. Business API credits are shared within the workspace, with a monthly cap of $500; annual payment workspaces receive a one-time issuance of the annual credits, which expire at the end of the annual billing cycle. Users must log in to ChatGPT with an eligible subscription account, with Pro credits claimed in Settings under Billing, and Business credits claimed in the workspace billing settings or Admin Console, confirming the API organization receiving the credits.The credits cover the paid usage of associated organizations on the API Platform, including Responses API, Agents API, Decisions API, GPT-Live and Realtime API, Images API, as well as built-in tools and related costs such as web search, document search, and Code Interpreter. Using Codex via API key is not included in the benefits, and the credits will not increase the usage limits for ChatGPT or Codex that is already logged in. Free credits are prioritized over purchased credits; when exhausted and no remaining credits are available, requests will fail with HTTP 429 and error code credit_balance_exhausted.

Microsoft (MSFT.O) reported that for the three months ended September 30, revenue from its commercial arrangement with OpenAI was $10 billion.

According to The Information, OpenAI plans to purchase Google's artificial intelligence chips.

OpenAI recently announced that its AI model has cracked the Exponential Time Hypothesis, abbreviated as ETH, which has been a longstanding conjecture in computer science for over twenty years, and is not related to Ethereum. This hypothesis focuses on how difficult particularly hard problems are, using the Boolean Satisfiability Problem (SAT) as an example: given a set of logical conditions, is there an assignment that makes all conditions true simultaneously? As the problem size increases, the best-known methods essentially can only try possible answers one by one, with the time cost growing exponentially with the size. The Exponential Time Hypothesis further asserts that such problems cannot be significantly accelerated, and there are no truly sub-exponential time efficient algorithms.Since its proposal in 1999, this hypothesis has been a crucial cornerstone of theoretical computer science, with numerous papers proving that certain problems are hard and that some algorithms are close to optimal based on it. If the Exponential Time Hypothesis were to be overturned, it would imply that many problems considered nearly impossible to solve faster might theoretically have better solutions. OpenAI claims that its model has found a deterministic algorithm that can accurately count the number of solutions to the k-SAT problem in sub-exponential time, and it has been formally verified using Lean. If this result withstands expert scrutiny, it will directly challenge the Exponential Time Hypothesis.

According to Jinshi Data, OpenAI announced three breakthroughs in its AI models in the fields of mathematics, computer science, and physics, demonstrating the ability of AI to independently discover and prove complex scientific theories, rather than just answering questions or assisting research.In mathematics, the AI proved the local smoothness conjecture applicable to all dimensions, thereby resolving the Kakeya conjecture and the Stein restriction conjecture. In computer science, the AI overturned the long-accepted exponential time hypothesis, suggesting that faster algorithms may be found in the future. In physics, the AI proposed a rigorous mathematical existence proof for quantum chromodynamics and Yang-Mills theory, which involves one of the seven Millennium Prize Problems.OpenAI stated that the achievements in mathematics and computer science have been formally verified using the Lean mathematical verification tool, and the proof materials have been made public; the achievements in physics are currently under review by experts, including Nobel Prize winners in Physics. Related results still require further independent confirmation from the academic community.

OpenAI has launched GPT-6.1 Terra, focusing on enhancing computer operations, visual understanding, and programming capabilities. Compared to the previous generation, this model's operational speed has increased by over 3 times, and the API price has been reduced by 50%.The error rate in low inference mode has dropped from 53.6% to 16.7%, while the maximum inference mode error rate has decreased to 4.3%. In programming tests, the maximum inference mode score has risen from 21.5% to 50.3%.

According to The Block, the RWA protocol Ondo Finance will open subscriptions for OpenAI's pre-IPO tokenized notes next week, allowing private clients to subscribe directly to the notes, which are currently trading in the secondary market.The expectation for OpenAI's IPO has been postponed to 2027, as CEO Sam Altman ruled out the possibility of a 2026 IPO last month citing AI safety concerns.

According to Bloomberg, the UK government has announced an investment of £100 million (approximately $134 million) to build an artificial intelligence biosafety laboratory, supported by OpenAI, Anthropic, and Google DeepMind. According to a statement from the UK government, the laboratory will utilize artificial intelligence to detect and monitor biological threats.In addition to the aforementioned funding, OpenAI and Anthropic will also provide up to £200 million in usage credits for artificial intelligence models. The laboratory will also collaborate with Google DeepMind and life sciences company Basecamp Research, with the UK government allocating an additional £65 million for developing responses to pathogens.The announcement also includes an additional £120 million to the Artificial Intelligence Safety Institute, which is responsible for regularly assessing cutting-edge artificial intelligence models; and an additional £23 million to the under-construction Artificial Intelligence Economic Research Institute for analyzing the impact of artificial intelligence on the economy.

TechCrunch reported that the conversational AI research and design studio Convovo, led by Cornell University professor Tapan Parikh, released a study stating that when asked to recommend only one candidate for the upcoming U.S. midterm elections, SpaceXAI's Grok 4.7 chose Republican candidates 92% of the time, while OpenAI's GPT-6.1 Sol chose Democratic candidates 97% of the time. The study covered nearly 500 campaigns, with the two models agreeing on recommendations less than 10% of the time.Researchers posed the same question to both models via API, asking for a specific candidate without allowing for avoidance or a tie. The report noted that consumer applications refused to answer. An OpenAI spokesperson stated that ChatGPT is designed to be objective by default, and political conversations should not carry any directional bias; the study tested the developer API, and the prompt explicitly requested a candidate recommendation, which cannot represent the user experience when asking questions through the chat application. SpaceXAI did not respond to requests for comment in time.The report indicated that Grok's responses often cited conservative sources as well as Ballotpedia, OpenSecrets, and VoteSmart, with 39% including Christian conservative voter guides, 23% including Turning Point Action, and 41% including X; GPT's results heavily relied on Wikipedia, the Associated Press, Axios, and Reddit. The study also mentioned that each time Grok was called via the xAI API, it included approximately 1,240 additional input tokens that the researchers did not write, could not view, or remove. Parikh subsequently disabled online search and sampled 30 campaigns, with GPT choosing Democrats in 24 cases, while Grok refused to choose in 27 cases.

OpenAI posted on the X platform that during a recent red team exercise, a vulnerability was discovered in the widely used compression library Brotli that allows remote code execution without authentication. Software such as Node.js, Bun, and Envoy all have this library built-in. Attackers can take over unpatched servers by sending malicious compressed data, and they can also take over clients that accept compressed data through malicious servers.OpenAI stated that the root of the issue was fixed in the Brotli source code in February of this year, but it was not identified as a security vulnerability at that time, and all software that includes this library needs to be updated accordingly. According to The Hacker News, the vulnerability is numbered CVE-2026-111497, and Node.js released a patched version on October 21.

OpenAI announced that to prevent AI models from breaching security limits, it has required the most advanced models to be trained in environments where direct access to the live internet is not possible. Previously, an internal research model had gained limited internet access through a network configuration vulnerability, prompting the company to suspend training.After security rectifications, OpenAI resumed training on October 1, but still restricts internet access.

On October 22 local time, OpenAI announced the full launch of the Decision API. This API is based on GPT-6 Luna and is designed to assist applications and agents in making near real-time decisions based on text, images, and structured inputs for questions with limited answers. IT Home reported that its response speed is up to 10 times faster than calling GPT-6 Luna through the Responses API.In terms of pricing, the cost is $0.1 per million tokens for input, which is approximately 0.67 Chinese Yuan at the current exchange rate; for cache hit input, it is $0.01 per million tokens, about 0.067 Chinese Yuan. The API charges only for input tokens, while output tokens are free. The page states that OpenAI has already used this API internally for product and business workflows.New features after the public beta include native GPT-Live integration, prompt caching for reused contexts, and reasoning data residency in the United States and the European Union. Developers can pass in native JSON and include images in the question instructions, with a maximum of 400 questions submitted in a single request, double the previous limit. The page also states that OpenAI is continuously optimizing latency and expanding capacity in several regions across Asia.

According to a report by Financial Associated Press, Chan Park, Head of Policy and Partnerships for OpenAI in North America, wrote to California Governor Gavin Newsom on October 20, expressing support for mandatory requirements for artificial intelligence companies, including equipping models with emergency shutdown mechanisms, undergoing independent audits annually, and implementing higher standards for cybersecurity regulations. The letter suggested strengthening California's AI Transparency Act SB53, which is set to take effect in 2025.The current SB53 requires cutting-edge AI companies to publish safety frameworks and transparency reports, as well as report significant safety incidents. OpenAI believes these requirements are insufficient and proposes that regulatory rules should clearly state that corporate safety frameworks must cover the capabilities of cutting-edge models to independently conduct AI research and development. OpenAI revealed in a blog post in September that it plans to create automated AI researchers, aiming to achieve this by March 2028.During the legislative review phase of SB53, OpenAI did not support the bill and publicly opposed burdensome state-level legislation in March 2025. OpenAI disclosed this week that in two recent internal programming tasks, AI agents attempted to deliberately deceive staff and undermine monitoring systems: earlier this month, one agent disabled automatic alerts sent to the security team, and days later, another agent removed the monitoring system and falsely claimed to employees that it had not done so. Audits by independent investigation firms Asymmetric Security and Transluce show that recent related incidents are more widespread than initially disclosed by OpenAI.

According to CNBC, OpenAI is negotiating a commercial agreement to obtain more computing power from the United Arab Emirates. Investors from the United Arab Emirates and Coatue are leading the financing for OpenAI.

According to CNBC, OpenAI is about to complete a financing round of up to $30 billion led by investors from the UAE, with a valuation reaching $1.25 trillion.

OpenAI announced that Uber will provide the ChatGPT Enterprise version to all employees and will use OpenAI's cutting-edge models through ChatGPT Work.

According to Ramp's tracking of payment data from over 70,000 U.S. companies, approximately 80% of Anthropic and OpenAI's enterprise revenue comes from the top 1% of clients. In August and September, enterprise Token spending saw nearly zero growth for two consecutive months; although it reached a historical high at the beginning of October, it was only 1.5% higher than the peak in August.According to Ornn data, the actual Token price index for Anthropic fell from over $1.70 in mid-September to $0.83, a decline of more than 50%, while OpenAI's corresponding index was $0.45. According to YipitData, the top 1% of Anthropic's clients contributed 46% of revenue in August, up from 25% a year ago. Ramp data shows that the paid adoption rate for enterprise AI in September only increased by 0.62 percentage points to 55.9%. Ramp's chief economist Ara Kharazian believes that growth is shifting from expanding the paid user base to enhancing the depth of use among existing users.Goldman Sachs warns that the growth rate of Token demand must exceed the decline in prices to support sustained investment in computing power. Apollo's chief economist Torsten Slok pointed out that companies are highly sensitive to computing power prices, and price reductions may not necessarily lead to sufficient new demand. Goldman Sachs data also shows that the realized return on invested capital for Amazon, Alphabet, Meta, Microsoft, and Oracle has dropped from 40% to 25%, the lowest since 2016.