OpenAI GPT Models Launch, Nvidia Chip News, Microsoft Cuts

OpenAI is making waves with the release of two new 'open-weight' AI models, gpt-oss-120b and gpt-oss-20b, designed for reasoning tasks. These models, available for free download, mark OpenAI's first open language model release in over five years, with CEO Sam Altman expressing the company's aim to promote AI development based on democratic values and ensure AI benefits all of humanity. The larger model can run on an Nvidia GPU, while the smaller one is designed to run on a laptop. However, the Allen Institute for AI (AI2) argues that OpenAI needs to be more open with AI models, as true progress requires open data, training methods, and shared evaluations, contrasting OpenAI's approach with its own fully open models like OLMo. OpenAI hopes that releasing these models will encourage businesses to stick with its technology and potentially pay for its more powerful products. These models are also now available on Amazon Web Services (AWS). In related news, Nvidia denies that its AI chips have a 'kill switch' after China raised concerns. Meanwhile, Microsoft is making further job cuts in Washington state, laying off another 40 employees, bringing the total to 3,160 since May, as it continues to invest heavily in AI infrastructure, with planned investments exceeding $30 billion. The AI landscape is also seeing intense competition for talent, with Meta's CEO Mark Zuckerberg reportedly offering one AI expert, Andrew Tulloch, a package potentially worth $1.5 billion over six years. Elsewhere, Londonderry Township in Dauphin County is considering zoning changes for AI data centers, reflecting the growing infrastructure needs of the AI industry. MLPerf Storage v2.0 benchmark results emphasize the importance of storage performance for AI training systems, showing that tested systems can now handle about twice as many accelerators as before. Finally, experts caution against the hype that AI can easily transform businesses overnight, emphasizing that the real value of AI lies in helping people with specific tasks and improving efficiency.

Key Takeaways

• OpenAI released two new 'open-weight' AI models, gpt-oss-120b and gpt-oss-20b, for free download, targeting AI reasoning tasks.
• The Allen Institute for AI (AI2) calls for deeper standards of openness from OpenAI, including sharing training data and methods.
• Nvidia denies its AI chips have a 'kill switch' following concerns raised by China.
• Microsoft is laying off an additional 40 employees in Washington state, bringing the total to 3,160 since May, while investing over $30 billion in AI infrastructure.
• Meta's CEO Mark Zuckerberg reportedly offered an AI expert a compensation package potentially worth $1.5 billion over six years.
• Londonderry Township is discussing zoning changes to accommodate the construction of AI data centers.
• MLPerf Storage v2.0 benchmark results highlight the critical role of storage performance in large-scale AI training.
• OpenAI's models are now available on Amazon Web Services (AWS).
• Data neutrality is key to AI's competitive advantage.
• Experts warn against overstating AI's immediate transformative capabilities, emphasizing its value in specific task automation.

OpenAI releases 'open' AI model but keeps secret tech hidden

OpenAI has launched its first open model since 2020, aiming to make AI more accessible. CEO Sam Altman hopes this release will encourage new research and product development. However, experts note that OpenAI is not revealing its most valuable technology like the architecture and training data behind models like GPT-5. The new models, gpt-oss-120b and gpt-oss-20b, target AI developers and can run on various hardware. OpenAI is responding to increasing competition in open-source AI.

OpenAI unveils two new 'open' AI models for reasoning tasks

OpenAI has launched two new AI models called gpt-oss-120b and gpt-oss-20b that are designed for AI reasoning. These models are free to download and use, and they are considered state-of-the-art for open models. The larger model, gpt-oss-120b, can run on a single Nvidia GPU, while the smaller one, gpt-oss-20b, can run on a laptop. This marks OpenAI's first open language model release in over five years. CEO Sam Altman said OpenAI wants to promote AI development based on democratic values.

OpenAI releases two new 'open' AI models for reasoning

OpenAI has released two new AI models, gpt-oss-120b and gpt-oss-20b, designed for reasoning tasks. These models are available for free download and are considered advanced for open models. The larger model can run on an Nvidia GPU, while the smaller one can run on a laptop. This is OpenAI's first open language model in over five years. CEO Sam Altman said OpenAI aims to ensure AI benefits all of humanity.

AI Institute says OpenAI needs to be more open with AI models

The Allen Institute for AI (Ai2) says OpenAI's new 'open-weight' models aren't truly open. Ai2 believes real progress in AI needs open data, training methods, and shared evaluations. OpenAI released details about its GPT-4 Turbo models, but didn't share its training data. Ai2 contrasts this with its own fully open models like OLMo. Ai2 calls for deeper standards of openness to boost collaboration and trust in AI.

OpenAI releases free AI models to compete with rivals

OpenAI has released two new AI models, gpt-oss-120b and gpt-oss-20b, that users can download and modify for free. This move is to compete with similar models from US and Chinese companies. These models are called 'open-weight language models' because their trained parameters are public. Meta and DeepSeek have similar open-source AI models. OpenAI says the new models are designed for AI tasks like web searching and coding, and can run on local computers.

Microsoft cuts 40 more jobs in Washington amid AI investments

Microsoft is laying off another 40 employees in Washington state. This brings the total job cuts to 3,160 since May. These cuts are separate from the larger layoffs announced earlier. Microsoft says these changes are a normal part of managing the business. The company is investing heavily in AI infrastructure, spending over $30 billion.

Microsoft cuts 40 Washington jobs while investing big in AI

Microsoft is laying off 40 more employees in Washington state as it increases its investments in AI. So far in 2025, the company has cut 3,160 jobs in Washington. Microsoft says these job cuts are part of necessary organizational changes. The company is making record profits from its cloud and AI business. Microsoft plans to invest an additional $30 billion in AI infrastructure.

AI training needs storage performance says MLPerf benchmark results

MLPerf Storage v2.0 benchmark results show that storage performance is important for AI training systems. The new checkpoint benchmarks give information for optimizing AI training. The benchmark results show that storage systems are improving quickly. Tested systems can now handle about twice as many accelerators as before. The v2.0 benchmark adds new tests that copy real-world checkpointing for AI training systems.

MLPerf highlights storage's key role in large-scale AI training

MLPerf Storage v2.0 benchmark results show that storage systems are important for AI training. The tests measure how well storage systems perform with machine learning tasks. The new benchmark includes tests that copy real-world checkpointing for AI training systems. The results show that storage systems are improving, and can now support twice as many accelerators. Stakeholders need to consider storage system performance to avoid slowing down AI training.

Londonderry Township to discuss zoning for AI data centers

Londonderry Township in Dauphin County is holding a public hearing to discuss zoning changes for AI data centers. Several plans are in progress to build AI data centers in Pennsylvania. The township is considering changes to zoning rules to allow data centers and their related infrastructure. A key area of interest is a property currently occupied by Seager Bros on East Harrisburg Pike. The meeting will be held at the Londonderry Township Volunteer Fire Department.

AI talent war offers billion-dollar jobs to top experts

There's a big competition for top AI experts, with some being offered huge amounts of money. Meta's CEO Mark Zuckerberg offered one AI expert, Andrew Tulloch, a package potentially worth $1.5 billion over six years. Zuckerberg is trying to attract talent from AI startups and rivals like Apple. However, many top researchers are driven more by a sense of purpose than by money. OpenAI's CEO Sam Altman believes that people who are passionate about their mission will be more successful than those motivated only by money.

AI's future is bright but not here yet warns expert

A dangerous idea is spreading that AI can easily transform businesses overnight, but that's not true yet. Reports paint a picture of AI agents running entire businesses, but this is disconnected from reality. This hype can be harmful and lead to disappointment. The real value of AI is in helping people with specific tasks, not in full automation. AI can improve efficiency and customer satisfaction by automating repetitive tasks.

OpenAI shares AI models to compete with other companies

OpenAI is now sharing two of its AI models used for online chatbots. This move is to compete with other companies that are sharing their technology through open source. OpenAI hopes this will encourage businesses to stick with its technology. The models, gpt-oss-120b and gpt-oss-20b, are not as powerful as OpenAI's top AI, but still rank among the world's leading models. OpenAI hopes that people using these models will also pay for its more powerful products.

Data neutrality is key to AI's competitive advantage

Data neutrality is now very important for AI companies to stay competitive. Companies need to control their data, models, and AI development without outside interference. The AI industry is changing with new partnerships and mergers. This creates risks for companies that have invested in their own AI models. Data neutrality helps companies keep their unique insights and control their valuable data.

OpenAI models now available on Amazon Web Services

OpenAI models are now available on Amazon Web Services (AWS) for the first time.

Nvidia denies its AI chips have a 'kill switch'

Nvidia says its AI chips do not have a 'kill switch' after China accused them of having one. Nvidia's Chief Security Officer David Reber said that their GPUs should not have kill switches or backdoors. This statement comes after China's Cyberspace Administration asked Nvidia for documents about security issues in the H20 chip. The US has placed export controls on some Nvidia chips to China for national security reasons. Nvidia CEO Jensen Huang believes it's better for the US if Nvidia's chips are the global standard.

Sources

• OpenAI’s new 'open' models are here—just don’t expect them to reveal the secret recipe
• OpenAI launches two ‘open’ AI reasoning models
• OpenAI launches two 'open' AI reasoning models
• ‘Open-weight’ debate: Allen Institute for AI says OpenAI needs to go further to be truly open
• OpenAI releases free, downloadable models in competition catch-up
• Microsoft cuts another 40 jobs in Washington state, continuing layoffs amid AI investment surge
• Microsoft lays off 40 more employees in Washington amid massive AI investment
• New MLPerf Storage v2.0 Benchmark Results Demonstrate the Critical Role of Storage Performance in AI Training Systems
• MLPerf Storage v2.0 Results Highlight Storage’s Role in AI Training at Scale
• Londonderry Township hosting public hearing to discuss potential zoning changes for AI data centers
• Inside AI's billion-dollar job offer lottery
• Don’t believe the hype: AI is the future, but we’re not there yet
• OpenAI to Open-Source Some of the A.I. Systems Behind ChatGPT
• Data neutrality: Safeguarding your AI's competitive edge
• For the first time, OpenAI models are available on AWS
• Nvidia says its AI chips don't have a 'kill switch' after Chinese accusation