DeepSeek R1 update proves that this is a positive threat to Openai and Google
This week, DeepSeek released an updated version of the R1 model on Huggingface, reigniting the open source and closed source competition. The updated version is called Deekseek-R1-0528,,,,, There are 685 billion parameters, from January version,,,,, There are 671 billion.
Unlike Openai and Google models, it is closed source, and the model weights of DeepSeek are publicly available. According to the benchmark, the R1-0528 update improves inference and reasoning capabilities and closes the gap with OpenAI’s O3 and Google’s Gemini 2.5 Pro.
DeepSeek also launched a distilled version of R1-0528 using Alibaba’s Qwen3 8b model. This is an example of a lightweight model that is less functional but also requires less computing power. DeepSeek-R1-0528-QWEN3-8B outperforms Google’s latest lightweight models Gemini-2.5-flash-inking-0520 and Openai’s O3-Mini in some benchmarks. But the bigger thing is that Deekseek’s distillation type is reported to be Running on a GPUaccording to TechCrunch.
To … refine all this information, Chinese competitors are taking an open approach that is cheaper and easier to access to catch up with U.S. competitors. Additionally, DeepSeek continues to prove that AI models may not require the computing power of OpenAI, Google, and other AI heavyweights. It can be said that looking at this space.
In other words, DeepSeek’s model also has its own shortcomings. According to an AI developer (by TechCrunch), the new DeepSeek update is More review Better than it Previous version Criticism about the Chinese government.
This tweet is currently unavailable. It may be loading or being deleted.
Of course, more has happened in the AI world over the past few days. This week’s product and feature news is easier after attending an AI event with Google, Anthropic and Microsoft last week.
This is one of the reasons DeepSeek’s R1 update has attracted the attention of the AI world this week. In other AI news, anthropomorphism finally gains voice mode, AI influencers will be popular, and the CEO of anthropomorphism warns that massive layoffs and AI-generated kangaroo.
Google’s VEO 3 swept the internet
On nearly every social media platform, users have made a new VEO 3 on Google’s new AI video model. The results are impressive and we’ve seen short films made entirely by VEO 3. Not bad for the product released 11 days ago.
This tweet is currently unavailable. It may be loading or being deleted.
This tweet is currently unavailable. It may be loading or being deleted.
This tweet is currently unavailable. It may be loading or being deleted.
No longer by AI video artist, a journalist from Wall Street Journal Using veo 3 to create a short film about yourself and robots.
Mashable’s tech editor Timothy Werth took a big week at Veo and came to a simple conclusion: we’re familiar with it.
More AI Product News: Cloud’s New Voice Mode and the Beginning of Age of Proxy Browsers
The amount of AI news this week is lighter after last week’s barrage. But what Once was Announcement this week is just as important.
Mixable light speed
Humans finally introduced their own Claude Voice Mode Compete with Chatgpt, Grok and Gemini. The feature is currently beta on the Claude app’s Mobile and even offers a plan of 20 to 30 voice conversations a day for free. Humans say you can ask Claude to sum up your calendar or read the document aloud. Payment subscribers can connect to Google Workspace for calendar, Gmail, and Docs access.
This tweet is currently unavailable. It may be loading or being deleted.
Openai is exploring capabilities Log in to third-party applications with chatgpt. We don’t know yet, but the company has posted an interest sheet on its website for developers to add this feature to their own apps using their engineering agent Codex. This may not sound like a big deal, but it basically means that users can easily link their personalized ChatGpt memories and settings to third-party apps, just like how you work when you log into a new app with your Google account.
Opera announced a new Agent AI browser called neon light. “It’s not just a place to view web pages, Neon can browse with you or for you, take action and help you get the job done,” the announcement reads. This includes the chatbot interface in the browser and the ability to fill in web forms for tasks such as booking and shopping. The announcement includes a promotional videotape that includes information about the humanoid robot browsing robot, which has little detail but says Neon will be a “premium subscription product” and there is a waitlist to register.
This tweet is currently unavailable. It may be loading or being deleted.
Now, browsers suddenly become a new field of AgentIC AI because of its ability to automate web search tasks. Confused is using a similar tool called cometand browser companies hub From its arc browser to an AI-centric browser called DIA. When Google might be Forced to sell chromiumwhich one Openai is friendly Take off your hand.
Dario Amodei’s predictions about entry-level jobs for AI replacement have begun to happen
Human CEO Dario Amodei warns In an interview with Axios, AI can “eliminate half of all entry-level white-collar jobs.”
Amodei’s forecast may be discovered as a new study by VC company SignalFire finds Hiring entry-level jobs has been reduced From 25% to 7% of the previous year. Some of these are due to changes in the economic environment, but AI is definitely a factor because companies choose to automate the less technical aspects of the work that new employees could have undertaken.
The latest culture of AI culture: AI-generated Kangaroo, Judge Judy and everything else
Google wants you to know its AI overview reaches 1.5 billion people per month. They may not want you to know that the AI overview is still struggling to count, spell and know this year. As Mashable’s Tim Marcin said, will AI Overview pass a Concussion Protocol?
This tweet is currently unavailable. It may be loading or being deleted.
The proposal for a 10-year ban on states that regulate AI is Very unpopularaccording to common sense media polls. The survey found that 57% of respondents opposed the suspension, including half of Republican respondents. As Mashable’s Rebecca Ruiz reported: “The vast majority of respondents, regardless of their political affiliation, agreed that Congress should not prohibit states from enacting or enforcing their own youth online security and privacy laws.”
In the private sector, The New York Times sign License Agreement with Amazon To allow its editing content to be used in Amazon’s AI models. The details are not clear, but from the outside, it seems to be a change era, at present Prosecuted Openai Copyrights allegedly infringe upon the use of its content to train its models.
That viral video about emotional support kangaroo holding a ticket and being denied boarding? This is generated by AIcertainly. A little obvious, but equally creepy, is another viral trend to use artificial intelligence to turn public figures like Emmanuel Macron and Justice Judy into public figures. Enter the baby. These are the weird AI-Slop intrusion times we live in.
This tweet is currently unavailable. It may be loading or being deleted.
AI also has some positive uses. This week, we learned a new humanoid robot from Huggingface (with the help of engineering from Robot Studios) that can be sold later this year for $3,000.
This tweet is currently unavailable. It may be loading or being deleted.
To end this review, a huge non-profit foundation has been developed AI Algorithm Detect the bird’s cry of the almost extinct teeth-mouthed pigeon. The teeth-mouthed pigeons are also known as the “little dodo” are Samoan ethnic birds, and scientists are using bioacoustic algorithms to locate and protect them.
Want to get the latest AI news from new product features to virus trends? Come back next week to check out another AI news Recap, meanwhile, follow @cecily_mauran and @mashable About more news.
Disclosure: Mashable’s parent company Ziff Davis filed a lawsuit against Openai in April, accusing it of infringing on Ziff Davis’ copyright in training and operating its AI systems.
