Latest news with #o3-mini
Express Tribune
29-04-2025
- Business
- Express Tribune
Alibaba launches Qwen3 'hybrid' AI model, challenges OpenAI, Google
Listen to article Alibaba Group on Monday unveiled Qwen3, a new family of large language models designed to compete with leading AI systems from OpenAI and Google. The release includes eight models, ranging from 0.6 billion to 235 billion parameters, and features a combination of dense and mixture-of-experts (MoE) architectures. The Chinese tech giant claims Qwen3 matches or outperforms the capabilities of OpenAI's o3-mini and Google's Gemini 2.5 Pro in several key benchmarks, including coding, math reasoning, and complex problem-solving. Qwen3 models support 119 languages and were trained on 36 trillion tokens sourced from textbooks, code, question-answer datasets, and AI-generated material. Unlike some competitors, Alibaba has made several Qwen3 models open-weight, available for download via Hugging Face and GitHub. The flagship Qwen-3-235B-A22B model, which achieved the highest scores in tests, remains restricted for now. Qwen3 introduces a 'hybrid reasoning' approach, allowing users to toggle between faster, non-reasoning outputs and slower, deeper reasoning modes to optimize accuracy. Alibaba says this flexibility enhances efficiency and user control over AI operations. The release intensifies competition in China's AI sector, following recent launches by DeepSeek and Baidu. It also comes amid heightened US export restrictions on advanced chips to China, which could impact future model training. Qwen3 will be available through cloud providers such as Fireworks AI and Hyperbolic, offering businesses new options beyond proprietary US AI systems. Industry observers say Alibaba's move signals a rapid closing of the gap between open and closed AI models globally. "Qwen3's performance shows that open models are keeping pace," said Tuhin Srivastava, CEO of AI platform Baseten. Alibaba previously released Qwen2.5-Max in January but says Qwen3 represents a significant leap in reasoning ability, coding proficiency, and multi-language support.
Yahoo
06-03-2025
- Business
- Yahoo
DuckDuckGo leans further into GenAI as its AI chat interface exits beta
Private search engine DuckDuckGo is leaning further into the generative AI opportunity. The non-tracking search engine has been dabbling with expanding the role of AI assistance in its product for the past year, including launching a chatbot-style interface last fall -- available at In a blog post Thursday, the company said the service is now exiting beta. It's also now simply called replacing the longer, mouthful name DuckDuckGo AI Chat. Users of can dip into AI models developed by the likes of Anthropic, OpenAI and Meta via a chatbot-style interface that sees their search queries handled in a conversational style. In other words, they get AI-generated answers to search asks powered by cutting-edge AI models that DuckDuckGo is making available, instead of the conventional search engine list of hyperlinks. DuckDuckGo notes that it has expanded the models users of can tap into -- with recent additions including OpenAI's o3-mini, Meta's Llama 3.3, and Mistral's Small 3. While access to is currently free, there is a daily limit on queries -- and DuckDuckGo says it is "exploring a paid plan for access to higher limits and more advanced (and costly) chat models." Simultaneously, the company is dialling up its use of GenAI in its conventional search engine interface -- at or -- by expanding the frequency that the search engine shows AI-assisted answers in response to a query. These are generative AI text summaries that can appear in response to search queries, above the usual blue links. "We now serve millions of AI-assisted answers daily. If you opt to show them often in our traditional search results, they should appear over 20% of the time," the company writes. Despite deepening its embrace of GenAI, the search firm is being careful to ensure that users retain agency over how much AI babble ends up in their search results. Users are able to choose the frequency that AI-generated answers will appear -- with "sometimes" being the default, and other options "often", "on-demand", and "never" letting users choose their own level of AI adventure. DuckDuckGo is responding to wider market shifts, as GenAI has continued to upend digital business-as-usual generally and web search specifically. Search kingpin Google has scrambled to fast-follow the viral fallout from OpenAI's AI chatbot, ChatGPT, and now embeds generative responses from its own AI models into search results. More recently, Google has even been experimenting with ditching links entirely in favor of AI summaries with a so-called AI Mode. So what can DDG bring to this competitive frenzy? The company clearly feels its core privacy pledge can transfer into this area. It offers users the chance to tap into major GenAI tools with reduced privacy risks, since they do not need to sign up for an account with an AI giant to get access. " allows you to use models from leading model providers without being tracked," its privacy policy suggests. "Chats are anonymized via proxying and never used for AI model training," DuckDuckGo also writes in the blog post, which touts the free (and sign-up free) access to what it bills as "private, useful, and optional AI." So the pitch here kind of boils down to 'have your GenAI cake and eat it in secret.' Although, if you read the full privacy policy, DuckDuckGo is careful to point out that if your search queries include your own personal data, that could end up sitting on the servers of large language model makers. Albeit, it stipulates, it's not tied back to your digital ID since it's masking IPs etc. Another feature DuckDuckGo is offering to make it easier to integrate GenAI into users' search workflows is the ability to easily switch between its conventional search engine interface to AI chat and vice versa. A chat button displayed below the search box of its search engine (see screengrab below) instantly transports the user to the AI chat interface -- with the prompt-field there pre-filled with whatever they had just been searching for on DDG's search engine, making it easy to hit the send button and get an AI-generated answer to the same query. Reversing this flow just requires tapping on the same (now highlighted) chat button to flip back to its conventional web search interface. So this is a best of both worlds approach to tapping GenAI in search. "We're finding that some people prefer to start in chat mode and then jump into more traditional search results when needed, while others prefer the opposite," DuckDuckGo writes, suggesting "some questions just lend themselves more naturally to one mode or the other, too." "So, we thought the best thing to do was offer both. We made it easy to move between them, and we included an off switch for those who'd like to avoid AI altogether," it adds. (Note: The chat button is also described as "optional" -- indicating that users can delve into settings to turn this off too if they don't even want a visual nudge towards GenAI.) While DuckDuckGo started with just Wikipedia as the source for its AI-assisted answers, it has since expanded to include sources from across the web, providing what it dubs as "prominent source links" with more information on what's underpinning the AI answers. Another update is a 'Recent Chats' feature that stores users' conversations with the AI search interface "locally on your device -- not on DuckDuckGo or other remote servers." Here, too, users can opt to disable the storage if they don't want any record of their chats kept at all, even on their own device. " chats are not used for any AI training, either by us or the underlying model providers," DuckDuckGo also writes. "To respond with answers and ensure all systems are working, these providers may store chats temporarily, but we remove all the metadata so there's no way for them to tie chats back to you personally." "On top of that, we have agreements in place with all providers to ensure that any saved chats are completely deleted within 30 days." Sign in to access your portfolio
Yahoo
05-03-2025
- Business
- Yahoo
Amazon is reportedly developing its own AI 'reasoning' model
Amazon reportedly wants to get in on the AI "reasoning" model game. According to Business Insider, Amazon is developing an AI model that incorporates advanced "reasoning" capabilities, similar to models like OpenAI's o3-mini and Chinese AI lab DeepSeek's R1. The model may launch as soon as June under Amazon's Nova brand, which the company introduced at its re:Invent developer conference last year. Reasoning models take a step-by-step, more considered approach to answering queries. This tends to boost their reliability in domains like math and science. The report says Amazon aims to adopt a "hybrid" reasoning architecture for its new model, along the lines of Anthropic's recently released Claude 3.7 Sonnet. Should that come to pass, the model could provide quick answers and more complex extended thinking within a single system. Amazon also hopes to make its Nova reasoning model more price-efficient than competitors, Business Insider claims. That might be a tall order. DeepSeek has developed a reputation for pricing its models incredibly cheaply.
Yahoo
27-02-2025
- Business
- Yahoo
OpenAI's new GPT-4.5 model is a better, more natural conversationalist
In what has already been a busy past few days for new model releases, OpenAI is capping off the week with a research preview of GPT-4.5. The company is touting the new system as its largest and best model for chat yet. In early testing, OpenAI says people found GPT-4.5 to be a more natural conversationalist, with the ability to convey warmth and display a kind of emotional intelligence. In one example shared by OpenAI, a person tells ChatGPT they're going through a hard time after failing a test. Where the company's previous models, including GPT-4o and o3-mini, might commiserate with the individual before offering a long list of unsolicited advice, GPT-4.5 takes a different tact. "Want to talk about what happened, or do you just need a distraction? I'm here either way," the chatbot says when powered by GPT-4.5. The gains shown by GPT-4.5 are the result of advancements OpenAI made in unsupervised learning. With unsupervised learning, a machine learning algorithm is given an unlabeled data set and left to its own devices to find patterns and insights. GPT-4.5 doesn't "think" like the company's state-of-the-art reasoning models, but in training the new model OpenAI made architectural enhancements and gave it access to more data and compute power. "The result is a model that has broader knowledge and a deeper understanding of the world, leading to reduced hallucinations," the company says. See for yourself — The Yodel is the go-to source for daily news, entertainment and feel-good stories. By signing up, you agree to our Terms and Privacy Policy. Speaking of reduced hallucinations, OpenAI measured how much better GPT-4.5 in that regard. When put through SimpleQA, an OpenAI-designed benchmark that tests large language models on their ability to answer "straightforward but challenging knowledge questions," GPT-4.5 beat out o3-mini, GPT-4o and even o1 with a hallucination rate of 37.1 percent. Obviously, the new model doesn't solve the problem of AI hallucinations altogether, but it is a step in the right direction. Despite its relative strengths over GPT-4o and o3-mini, GPT-4.5 isn't a direct replacement for those models. Compared to OpenAI's reasoning systems, GPT-4.5 is "a more general-purpose, innately smarter model." Additionally, it's not natively multimodal like GPT-4o, meaning it doesn't work with features like Voice Mode, video or screensharing. It's also "a very large and compute-intensive model." It's best to think of GPT-4.5 as a stepping stone to systems OpenAI plans to offer in the future. In fact, Sam Altman said as much earlier this month when he shared the company's roadmap, noting GPT-4.5 would be "our last non-chain-of-thought model" — referring to the fact that the new system doesn't solve problems by tackling them step by step like OpenAI's reasoning models do. Its successor, GPT-5, will likely integrate many of OpenAI's latest technologies, including its frontier o3 model. OpenAI reiterated that today, saying it plans to bring GPT-4.5's "unique strengths, including broader knowledge, stronger intuition, and greater 'EQ,' to all users in future models." In the meantime, ChatGPT Pro subscribers can begin using GPT-4.5 starting today, with Pro and Team users slated to gain access starting next week.
Yahoo
18-02-2025
- Business
- Yahoo
xAI launches Grok 3 AI, claiming it is capable of 'human reasoning'
xAI has launched its Grok 3 models during a livestream with Elon Musk, who said they were "an order of magnitude more capable than Grok 2." The Grok 3 mini model can answer questions quickly, but it's not as accurate as the other models in the family. Meanwhile, the Grok 3 Reasoning and Grok 3 mini Reasoning models are capable of mimicking human-like reasoning when it comes to analyzing information the user needs. Other examples of AI models capable of reasoning tasks are DeepSeek's R1 and OpenAI's o3-mini. According to TechCrunch, xAI claimed during the event that Grok 3 Reasoning performed better than the best version of o3-mini on several benchmarks. Grok 3's features will initially be available to subscribers paying for X's Premium+ tier, which now costs $40 a month in the US. (X raised the Premium+ tier's pricing from $16 to $22 in December — now, less than two months later, it's almost twice as expensive.) They will also be available through an upcoming separate subscription option for the standalone Grok app and Grok on the web. Based on leaked information, the subscription option will be called SuperGrok and will cost $30 a month. With the Grok 3 models enabled, users will be able to ask the chatbot to "Think" if they want to tap its reasoning capabilities for mathematics, science and programming questions. For even more complex queries, they can use the "Big Brain" function that requires additional computing. The models' reasoning capabilities power a new Grok feature called DeepSearch, which xAI describes as the "next generation search engine." DeepSearch will scan the internet and X, formerly Twitter, to conjure a brief summary for research inquiries. In addition to launching the Grok 3 models, xAI also revealed during the event that the Grok app will get a "voice mode" within a week, giving it synthesized voices to converse with users. Grok 2, the company's older models, will be open sourced in the coming months.
