Anthropic Launches the World's First ‘Hybrid Reasoning' AI Model
Feb 24, 2025 1:43 PM Claude 3.7, the latest model from Anthropic, can be instructed to engage in a specific amount of reasoning to solve hard problems.
Anthropic, an artificial intelligence company founded by exiles from OpenAI, has introduced the first AI model that can produce either conventional output or a controllable amount of 'reasoning' needed to solve more grueling problems.
Anthropic says the new hybrid model, called Claude 3.7, will make it easier for users and developers to tackle problems that require a mix of instinctive output and step-by-step cogitation. 'The [user] has a lot of control over the behavior—how long it thinks, and can trade reasoning and intelligence with time and budget,' says Michael Gerstenhaber, product lead, AI platform at Anthropic.
Claude 3.7 also features a new 'scratchpad' that reveals the model's reasoning process. A similar feature proved popular with the Chinese AI model DeepSeek. It can help a user understand how a model is working over a problem in order to modify or refine prompts.
Dianne Penn, product lead of research at Anthropic, says the scratchpad is even more helpful when combined with the ability to ratchet a model's 'reasoning' up and down. If, for example, the model struggles to break down a problem correctly, a user can ask it to spend more time working on it.
Frontier AI companies are increasingly focused on getting the models to 'reason' over problems as a way to increase their capabilities and broaden their usefulness. OpenAI, the company that kicked off the current AI boom with ChatGPT, was the first to offer a reasoning AI model, called o1, in September 2024. OpenAI has since introduced a more powerful version called o3, while rival Google has released a similar offering for its model Gemini, called Flash Thinking. In both cases, users have to switch between models to access the reasoning abilities—a key difference compared to Claude 3.7.
A user view of Claude 3.7 Courtesy of Anthropic
The difference between a conventional model and a reasoning one is similar to the two types of thinking described by the Nobel-prize-winning economist Michael Kahneman in the book 2011 Thinking Fast and Slow : fast and instinctive System-1 thinking; and slower more deliberative System-2 thinking.
The kind of model that made ChatGPT possible, known as an LLM, produces instantaneous responses to a prompt by querying a large neural network. These outputs can be strikingly clever and coherent but may fail to answer questions that require step-by-step reasoning, including simple arithmetic.
An LLM can be forced to mimic deliberative reasoning if it is instructed to come up with a plan that it must then follow. This trick is not always reliable, however, and models typically struggle to solve problems that require extensive careful planning. OpenAI, Google, and now Anthropic are all using a machine learning method known as reinforcement learning to get their latest models to learn to generate reasoning that points towards correct answers. This requires gathering additional training data from humans on solving specific problems.
Penn says that Claude's reasoning mode received additional data on business applications including writing and fixing code, using computers, and answering complex legal questions. 'The things that we made improvements on are [...] technical subjects or subjects which require long reasoning,' Penn says. 'What we have from our customers is a lot of interest in deploying our models into their actual workloads.'
Anthropic says that Claude 3.7 is especially good at solving coding problems that require step-by-step reasoning, outscoring OpenAI's o1 on some benchmarks like SWE-bench. The company is today releasing a new tool, called Claude Code, specifically designed for this kind of AI-assisted coding.
'The model is already good at coding,' Penn says. '[But] additional thinking would be good for cases that might require very complex planning—say you're looking at an extremely large code base for a company.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Tom's Guide
43 minutes ago
- Tom's Guide
I ignored this ChatGPT setting for months — now I use it every day
Like millions of others, I use ChatGPT daily. As a power user, I frequently use ChatGPT to summarize research, create images and I even use the bot to talk me down from a spiral of recently, I revisited an underused setting buried in the app that completely changes how the bot responds. It's so useful that I really wish I had taken advantage of it sooner. No, it's not a secret plugin or a pro-only feature. It's something that's been there the whole time: custom instructions. You've probably seen the button dozens of times but never bothered to click on it. It lives quietly in your settings menu under the heading: 'Customize ChatGPT.' This customizes the chatbot's behavior. Using customized GPTs dramatically improved how I interact with the AI by tailoring the chatbot to my specific of it as crafting perfect assistants for certain tasks. Instead of re-explaining your preferences every time, a custom GPT can be set up to understand your job, writing style, tone and even the kind of responses you want (short, long, complex, simple).Custom GPTs also support powerful tools and integrations. You can grant them access to a code interpreter, web browsing, image generation or even custom APIs and uploaded this feature turns ChatGPT into an even better assistant capable of analyzing data, generating visuals or referencing your documents without extra work on your part. Plus, you can set behavioral instructions so the GPTs always respond in your preferred tone or format, saving you time and improving consistency. For me, the hardest part about using a customized GPT is literally remembering to use it. Although it's just a click away, sometimes I'll dive into a prompt before I remember there's a better way to get the best you've built a custom GPT that works well, you can reuse it as often as you like or share it with others (a great asset for teams). Whether you're managing SEO, writing emails or brainstorming ideas, having a GPT fine-tuned to your process means faster, smarter output. It's a useful way to turn a general-purpose tool into a personal or professional super-assistant. If you're like me, you'll notice that when you finally start customizing GPTS, your experience with AI will shift entirely, and for the better. Your responses will feel clearer and far more personal. You won't get generic responses, but answers that fit your style and suggestions that are more engaging. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. You can access this feature in just a few taps: Click your name (or the three dots in the bottom left) Tap Settings Select Custom Instructions You'll see two key fields: 'What would you like ChatGPT to know about you to provide better responses?'(Example: 'I'm a busy mom of three and want an empathetic, conversational tone that feels like I'm chatting with a friend.' 'How would you like ChatGPT to respond?'(Example: 'Use short paragraphs, avoid buzzwords, and give practical suggestions. Add a human tone, like you're texting a smart coworker.') Once you fill these out, that context is baked into every conversation. You don't have to reintroduce yourself or explain your tone again. ChatGPT just gets it. Custom instructions are convenient because they are like having a hat trick in your back pocket. Whatever issue you were having with ChatGPT earlier, such as answers feeling too generic or formal, this setting fixes turns the chatbot into something much closer to a real assistant and one that actually understands everything about you (well, as much as you feel comfortably telling it). It also means you'll spend less time rewriting responses and more time getting useful results. For example: When I asked it to write a note to the babysitter, it used formatting and tone I'd normally have to adjust. When I needed a list of birthday party locations in the area, it knew where I lived and pulled them up immediately. (This type of ultra personalization might not be for everyone, but I find it to be a time saver). And when I asked for snack ideas for the soccer team, the list actually sounded like something I'd submit (and easy enough for a busy mom to contribute); not something from a generic listicle generator. Best of all? It feels more personal without sacrificing quality responses. If you're using ChatGPT with memory enabled, custom instructions are the perfect complement. Memory helps the chatbot remember ongoing preferences and facts across conversations, while custom instructions give it a solid starting point for every new chat. Even if memory isn't your thing, these static instructions make ChatGPT far more efficient right out of the gate. This one setting changes how well ChatGPT can work for you. If you've been using the chatbot like a search engine or idea machine, custom instructions push it into a new category making it more like a personal AI assistant. You'll notice a difference when it starts answering like someone who knows your voice, your goals and how you think. Do you use custom GPTs? Let me know in the comments!
Miami Herald
an hour ago
- Miami Herald
Trump's China gambit belies rocky road ahead on tariff deals
President Donald Trump has come up short on striking trade deals with most nations with just one month left before his self-imposed tariff deadline, even as he took his first steps in weeks toward engaging with China. Trump secured a much-desired call with Chinese President Xi Jinping, paving the way for a new round of talks on Monday in London - yet the diplomacy was overshadowed by a blowout public fight between Trump and his billionaire onetime ally, Elon Musk. Trump's aides insisted Friday that the president was moving on and focused on his economic agenda. Still, question marks remain over the U.S.'s most consequential trade relationships, with few tangible signs of progress toward interim agreements. India, which the Trump administration has cited as an early deal target, has taken a tougher line in negotiations and challenged Trump's auto tariffs at the World Trade Organization. Japan held another round of talks with the U.S., while also signaling it wants a reprieve from duties on cars and light trucks. The legal fight over Trump's tariffs hangs over everything. A court ruling striking down the country-by-country duties imposed using emergency authorities left partners with no certainty over what Trump's powers are. The next test could come as soon as next week, when a court could rule on the administration's appeal. Trump and his team were eager to draw attention to inroads with China as proof his ways are working. Trump on Friday described talks with Beijing as 'very far advanced' and said Xi had agreed to speed shipments of critical rare-earth minerals that were at the center of recent tension. Unlocking those supplies would spell relief for major American automakers. Chinese Vice Premier He Lifeng will visit the U.K. next week, during which he will conduct trade negotiations with the U.S., the Chinese foreign ministry said in a statement late Saturday. The mixed results in the talks so far demonstrate the highs and lows of Trump's mercurial approach to trade, in which he and aides have cast him as the ultimate decision-maker on any deals. Rather than provide a clear-cut victory, Trump's dealings with Xi also show the difficult road ahead with China. The rare-earths dispute revealed how important those supplies, which Beijing dominates, are for the U.S. economy. 'Xi is not letting go of the rare earths. He's got leverage, he's using it,' said Douglas Holtz-Eakin, president of the American Action Forum, a conservative think tank. 'They talked, that's the most important thing. I think they're really far apart.' The clock is ticking for Trump. His 90-day pause on higher tariffs for the European Union and nearly five dozen countries expires July 9 — barring an extension he could do with the flick of a pen — while China's reprieve extends until August. If deals aren't reached, Trump plans to restore tariff rates to the levels he first announced in April, or lower numbers that exceed the current 10% baseline, a White House official said, speaking on condition of anonymity. 'We will have deals. It takes time. Usually it takes months and years; in this administration, it's going to take more like days,' White House trade counselor Peter Navarro said Friday on Fox Business. 'We're on task and on target.' The Office of the U.S. Trade Representative 'looks more like a deli now,' Navarro said, with countries lining up for talks. USTR sent letters this week to trading partners reminding them of the deadline. It's unclear what all the frantic activity has yielded. Xi for months was reluctant to get on the phone with Trump and analysts speculated about what concessions the U.S. president offered to his counterpart in exchange for the call. Trump at least appeared to give some ground on foreign students, saying it would be his 'honor' to welcome Chinese scholars even as his administration cracks down on student visas. German Chancellor Friedrich Merz visited Washington facing demands from his nation's automakers for tariff credits for vehicles they produce in the U.S. But the subject barely came up during the public portion of his meeting with Trump, who spent a large chunk of time unloading on Musk. 'We'll end up hopefully with a trade deal or we'll do something - you know, we'll do the tariffs,' Trump said Thursday alongside Merz. Merz, in his U.S. visit, emphasized the integrated trade ties between countries that are at risk — including by personally driving a BMW built in South Carolina. The German leader said Friday at an industry event the nations should agree on an 'offset rule' that would provide tariff relief for existing U.S. production. Trump's U.K. deal - the lone pact so far — was undercut this week when he plowed ahead with levies on steel and aluminum. The U.K. said the pact included an agreement for zero tariffs on British metals, but Trump's latest order kept a 25% charge on them while negotiations continue and doubled the rate for others. Still, the upcoming Group of Seven summit of leaders from major economies could provide an opportunity for the type of in-person dealmaking Trump craves. Canadian Prime Minister Mark Carney has been discussing terms of a potential interim deal with Trump ahead of the gathering this month near Calgary. One theme is clear: Negotiations over his so-called reciprocal tariffs have grown intertwined with his separate duties on autos and metals, despite previous U.S. signals that the administration considered them separate. 'He's entirely transactional,' Holtz-Eakin said of Trump. 'He will always deal.' Talks are ongoing with the EU, which has previously proposed an agreement with the U.S. to mutually drop auto tariffs to zero as part of a broader trade framework, which the Trump administration rejected. The bloc subsequently suggested working toward zero-for-zero tariffs on cars, other industrial goods and some agricultural imports with tariff-rate quotas as a possible interim measure. Commerce Secretary Howard Lutnick said this week he'd consider some type of 'export credit' on autos, the kind of carve-out sought by Germany on vehicle tariffs. And he predicted there would be a U.S.-India deal in the 'not too distant future.' Lutnick signaled, though, Trump's push for so-called reciprocity comes with caveats. The U.S. wouldn't agree with Vietnam to drop all tariffs, because it believes the Southeast Asian nation is a hub for so-called transshipment of Chinese goods. Talks with South Korea, where Trump spoke with newly elected president Lee Jae-myung, and Japan, which had top trade negotiator Ryosei Akazawa meet with Lutnick, continued this week. In yet another sign of the Trump team's frenetic approach, Nikkei reported that different — and even competing — positions among Treasury Secretary Scott Bessent, Trade Representative Jamieson Greer and Lutnick had confounded Japanese counterparts. --- (Akayla Gardner, Jennifer A. Dlouhy, Alberto Nardelli, Hadriana Lowenkron, Arne Delfs and Shiyin Chen contributed to this report.) Copyright (C) 2025, Tribune Content Agency, LLC. Portions copyrighted by the respective providers.
Yahoo
3 hours ago
- Yahoo
Apple's Siri Could Be More Like ChatGPT. But Is That What You Want?
I've noticed a vibe shift in the appetite for AI on our devices. My social feeds are flooded with disgust over what's being created by Google's AI video generator tool, Veo 3. The unsettling realistic video of fake people and voices it creates makes it clear we will have a hard time telling apart fiction from reality. In other words, the AI slop is looking less sloppy. Meanwhile, the CEO of Anthropic is warning people that AI will wipe out half of all entry-level white-collar jobs. In an interview with Axios, Dario Amodei is suggesting government needs to step in to protect us from a mass elimination of jobs that can happen very rapidly. So as we gear up for Apple's big WWDC presentation on Monday, I have a different view of headlines highlighting Apple being behind in the AI race. I wonder, what exactly is the flavor of AI that people want or need right now? And will it really matter if Apple keeps waiting longer to push out it's long promised (and long delayed) personalized Siri when people are not feeling optimistic about AI's impact on our society? In this week's episode of One More Thing, which you can watch embedded above, I go over some of the recent reporting from Bloomberg that discusses leadership changes on the Siri team, and how there are different views in what consumers want out of Siri. Should Apple approach AI in a way to make Siri into a home-grown chatbot, or just make it a better interface for controlling devices? (Maybe a bit of both.) I expect a lot of griping after WWDC about the state of Siri and Apple's AI, with comparisons to other products like ChatGPT. But I hope we can use those gripes to voice what we really want in the next path for the assistant, by sharing our thoughts and speaking with our wallet. Do you want a Siri that's better at understanding context, or one that goes further and makes decisions for you? It's a question I'll be dwelling on more as Apple gives us the next peak into the future of iOS on Monday, and perhaps a glimpse of how the next Siri is shaping up. If you're looking for more One More Thing, subscribe to our YouTube page to catch Bridget Carey breaking down the latest Apple news and issues every Friday.
