Tech
DeepSeek: Everything you need to know about the AI chatbot app
DeepSeek has gone viral.
Chinese AI lab DeepSeek broke into the mainstream consciousness this week after its chatbot app rose to the top of the Apple App Store charts (and Google Play, as well). DeepSeek’s AI models, which were trained using compute-efficient techniques, have led Wall Street analysts — and technologists — to question whether the U.S. can maintain its lead in the AI race and whether the demand for AI chips will sustain.
But where did DeepSeek come from, and how did it rise to international fame so quickly?
DeepSeek’s trader origins
DeepSeek is backed by High-Flyer Capital Management, a Chinese quantitative hedge fund that uses AI to inform its trading decisions.
AI enthusiast Liang Wenfeng co-founded High-Flyer in 2015. Wenfeng, who reportedly began dabbling in trading while a student at Zhejiang University, launched High-Flyer Capital Management as a hedge fund in 2019 focused on developing and deploying AI algorithms.
In 2023, High-Flyer started DeepSeek as a lab dedicated to researching AI tools separate from its financial business. With High-Flyer as one of its investors, the lab spun off into its own company, also called DeepSeek.
From day one, DeepSeek built its own data center clusters for model training. But like other AI companies in China, DeepSeek has been affected by U.S. export bans on hardware. To train one of its more recent models, the company was forced to use Nvidia H800 chips, a less-powerful version of a chip, the H100, available to U.S. companies.
DeepSeek’s technical team is said to skew young. The company reportedly aggressively recruits doctorate AI researchers from top Chinese universities. DeepSeek also hires people without any computer science background to help its tech better understand a wide range of subjects, per The New York Times.
DeepSeek’s strong models
DeepSeek unveiled its first set of models — DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat — in November 2023. But it wasn’t until last spring, when the startup released its next-gen DeepSeek-V2 family of models, that the AI industry started to take notice.
DeepSeek-V2, a general-purpose text- and image-analyzing system, performed well in various AI benchmarks — and was far cheaper to run than comparable models at the time. It forced DeepSeek’s domestic competition, including ByteDance and Alibaba, to cut the usage prices for some of their models, and make others completely free.
DeepSeek-V3, launched in December 2024, only added to DeepSeek’s notoriety.
According to DeepSeek’s internal benchmark testing, DeepSeek V3 outperforms both downloadable, openly available models like Meta’s Llama and “closed” models that can only be accessed through an API, like OpenAI’s GPT-4o.
Equally impressive is DeepSeek’s R1 “reasoning” model. Released in January, DeepSeek claims R1 performs as well as OpenAI’s o1 model on key benchmarks.
Being a reasoning model, R1 effectively fact-checks itself, which helps it to avoid some of the pitfalls that normally trip up models. Reasoning models take a little longer — usually seconds to minutes longer — to arrive at solutions compared to a typical non-reasoning model. The upside is that they tend to be more reliable in domains such as physics, science, and math.
There is a downside to R1, DeepSeek V3, and DeepSeek’s other models, however. Being Chinese-developed AI, they’re subject to benchmarking by China’s internet regulator to ensure that its responses “embody core socialist values.” In DeepSeek’s chatbot app, for example, R1 won’t answer questions about Tiananmen Square or Taiwan’s autonomy.
A disruptive approach
If DeepSeek has a business model, it’s not clear what that model is, exactly. The company prices its products and services well below market value — and gives others away for free.
The way DeepSeek tells it, efficiency breakthroughs have enabled it to maintain extreme cost competitiveness. Some experts dispute the figures the company has supplied, however.
Whatever the case may be, developers have taken to DeepSeek’s models, which aren’t open source as the phrase is commonly understood but are available under permissive licenses that allow for commercial use. According to Clem Delangue, the CEO of Hugging Face, one of the platforms hosting DeepSeek’s models, developers on Hugging Face have created over 500 “derivative” models of R1 that have racked up 2.5 million downloads combined.
DeepSeek’s success against larger and more established rivals has been described as “upending AI” and “over-hyped.” The company’s success was at least in part responsible for causing Nvidia’s stock price to drop by 18% in January, and for eliciting a public response from OpenAI CEO Sam Altman.
Microsoft announced that DeepSeek is available on its Azure AI Foundry service, Microsoft’s platform that brings together AI services for enterprises under a single banner. When asked about DeepSeek’s impact on Meta’s AI spending during its first-quarter earnings call, CEO Mark Zuckerberg said spending on AI infrastructure will continue to be a “strategic advantage” for Meta.
During Nvidia’s fourth-quarter earnings call, CEO Jensen Huang emphasized DeepSeek’s “excellent innovation,” saying that it and other “reasoning” models are great for Nvidia because they need so much more compute.
At the same time, some companies are banning DeepSeek, and so are entire countries and governments, including South Korea. New York state also banned DeepSeek from being used on government devices.
As for what DeepSeek’s future might hold, it’s not clear. Improved models are a given. But the U.S. government appears to be growing wary of what it perceives as harmful foreign influence.
TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.
This story was originally published January 28, 2025, and will be updated regularly.
Tech
Exclusive: Google deepens Thinking Machines Lab ties with new multi-billion-dollar deal
Former OpenAI executive Mira Murati’s startup, Thinking Machines Lab, has signed a new multi-billion-dollar agreement to expand its use of Google Cloud’s AI infrastructure, including systems powered by Nvidia’s latest GPUs, TechCrunch has exclusively learned.
The deal is valued in the single-digit billions, according to a source familiar with the matter, and includes access to Google’s latest AI systems built atop Nvidia’s new GB300 chips, alongside infrastructure services to support model training and deployment.
Google has been actively striking a number of cloud deals with AI developers as it aims to wrap together its AI computing offerings with other cloud services like storage, a Kubernetes engine, and Spanner, its database product. Earlier this month, Anthropic signed an agreement with Google and Broadcom for multiple gigawatts of tensor processing unit (TPUs) capacity (these are Google’s custom-designed AI chips for machine learning workloads).
But the competition is fierce. Just this week, Anthropic also signed a new agreement with Amazon to secure up to 5 gigawatts of capacity for training and deploying Claude.
Earlier this year, Thinking Machines partnered with Nvidia in a deal that included an investment from the chipmaker. But this is the first time the lab has struck a deal with a cloud services provider. The deal is not exclusive, so Thinking Machines may use multiple cloud providers over time, but it’s still a sign that Google is looking to lock in fast-growing frontier labs early.
Murati left her job as OpenAI’s chief technologist and founded Thinking Machines in February 2025. The company, which soon afterwards raised a $2 billion seed round at a $12 billion valuation, has remained highly secretive, but launched its first product in October. Dubbed Tinker, it’s a tool that automates the creation of custom frontier AI models.
Wednesday’s deal provided some insight into what Thinking Machines is developing. In a press release, Google noted that it can support the startup’s reinforcement learning workloads, which Tinker’s architecture relies on. Reinforcement learning is a training approach that has underpinned recent breakthroughs at labs, including DeepMind and OpenAI, and the scale of the Google Cloud deal reflects how computationally expensive that work can get.
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
Thinking Machines is among the first Google Cloud customers to access its GB300-powered systems, which offer a 2X improvement in training and serving speed compared to prior-generation GPUs, per Google.
“Google Cloud got us running at record speed with the reliability we demand,” Myle Ott, a founding researcher at Thinking Machines, said in a statement.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
The most interesting startups showcased at Google Cloud Next 2026
Google Cloud Next is taking place this week in Las Vegas, and one clear message has emerged: Google wants AI startups on its cloud. To that end, it made several startup-related announcements.
The most significant is that the tech giant has earmarked a new $750 million budget to help its Cloud partners sell more AI agents to enterprises. This funding is available to partners ranging from startups to the big consulting firms. It can be used for costs like Gemini proof-of-concept projects, Google forward-deployed engineers, cloud credits, and deployment rebates.
Google also highlighted a long list of startups that are using Google Cloud, either newly signed or expanding their footprint. Among them are a few standout names:
Lovable is expanding its use of Google Cloud by launching a new coding agent through Google’s enterprise app marketplace. Lovable is the fast-growing vibe coding startup and was on a $400 million ARR track as of February, it said.
Notion, Silicon Valley’s favorite AI-infused document productivity app, most recently valued at about $11 billion, is using Gemini models to power its text and image generation features.
Gamma, an AI-powered PowerPoint killer recently valued at a $2.1 billion valuation, is using Google’s state-of-the-art image model Nano Banana 2 and other Google Cloud features.
Inferact, the commercial inference startup from the creators of the popular open-source project vLLM, is accessing Nvidia’s GPUs through Google Cloud, in addition to using the tech giant’s AI stack.
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
ComfyUI, the popular open-source tool for creating AI-generated images and multimedia, also offers access to Nano Banana 2 and is using other Cloud features.
Other startups that received the Google Cloud shout-out this year include:
ChorusView, which makes AI-powered smart tags that track the condition and movement of goods in real time.
Emergent AI, a vibe coding platform.
ExaCare AI, which makes AI software for post-acute medical care facilities.
Insilica, which creates AI-generated regulatory-compliant chemical safety reports.
Optii, which makes AI-enhanced hotel operations software.
Parallel AI, which builds web search and research APIs built for AI agents.
Proximal Health, which makes AI-powered software that automates the insurance claims adjudication process.
Reducto, which does AI-powered document parsing.
Stord, which handles e-commerce fulfillment and parcel operations.
Stylitics, which makes AI image generation software for retailers for tasks like outfit styling and product bundles.
Temporal, a developer cloud environment built to prevent failures.
Vapi, which makes dev tools for building conversational voice agents.
Vurvey Labs, which conducts synthetic market research via AI agents.
Wand, an in-game assistant for single-player PC games.
Watershed, which makes software that helps enterprises report on and manage sustainability programs.
ZenBusiness, an all-in-one back-office tool for small businesses that includes an AI chat assistant.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Duolingo is now giving free users access to advanced learning content
Duolingo announced on Wednesday that its advanced language learning content is now available for free across nine languages: English, Spanish, French, German, Italian, Portuguese, Japanese, Korean, and Chinese. Users can access this content through the web, iOS, and Android devices.
This advanced content is at the B2 level on the Common European Framework of Reference for Languages (CEFR), which is the international standard for language skills that schools and employers recognize. B2 level content refers to learning materials without translations, complex scenarios, and specialized vocabulary.
The new offering will include features like “Advanced Stories,” which helps with reading comprehension, and DuoRadio, a podcast-like audio experience for listening comprehension.
Now that Duolingo users can tap into this advanced learning content for free, they can level up their skills, whether that’s practicing for job interviews, prepping for studying abroad, or tackling complex news articles, films, and books without relying on translations.
The company says this positions it as the only free app to offer advanced-level learning across these nine languages at no cost. While competitors like Babbel and Busuu offer advanced courses, they typically require paid subscriptions. For instance, Busuu has some CEFR-aligned courses up to the B2 level, but the free version is pretty limited and doesn’t offer lessons like grammar explanations, so users need to pay for full access.
Previously, Duolingo only provided free courses that capped at A2 or B1 levels, mainly focusing on basic communication skills.
The company is positioning this free advanced learning offering as an enticing opportunity for job seekers, framing language learning as a practical pathway to improving employability in an increasingly global workforce.
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
This comes at a time when the job market remains highly competitive and overall growth has slowed. Research from the American Council on the Teaching of Foreign Languages shows that learning a second language can raise someone’s employability by as much as 50%.
“Reaching job-ready proficiency in a new language used to be out of reach for most people,” Bozena Pajak, head of learning science at Duolingo, said in a statement. “It took years of expensive classes or immersive experiences that not everyone could access.”
Duolingo’s decision to offer advanced learning for free is also a strategy to increase its free user base. In its Q4 earnings report, the company stated that it has 52.7 million daily active users, demonstrating 30% growth compared to the previous year. This number is higher than its paid subscriber base, which stands at 12.2 million. However, Duolingo’s shares fell after the company projected that the year-over-year bookings growth rate for Q2 2026 is expected to experience a slight decline.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Apples excellent AirPods Pro 3 are back on sale for just $200 — a $50 savings
Red Bulls' struggling defense clashes with woeful D.C. offense
