The AI-generated video scene has been hopping this year (or twirling wildly, as the case may be). This past week alone we've seen releases or announcements of OpenAI's Sora, Pika AI's Pika 2, Google's Veo 2, and Minimax's video-01-live. It's frankly hard to keep up, and even tougher to test them all. But recently, we put a new open-weights AI video synthesis model, Tencent's HunyuanVideo, to the test—and it's surprisingly capable for being a "free" model.
Unlike the aforementioned models, HunyuanVideo's neural network weights are openly distributed, which means they can be run locally under the right circumstances (people have already demonstrated it on a consumer 24 GB VRAM GPU) and it can be fine-tuned or used with LoRAs to teach it new concepts.
Notably, a few Chinese companies have been at the forefront of AI video for most of this year, and some experts speculate that the reason is less reticence to train on copyrighted materials, use images and names of famous celebrities, and incorporate some uncensored video sources. As we saw with Stable Diffusion 3's mangled release, including nudity or pornography in training data may allow these models achieve better results by providing more information about human bodies. HunyuanVideo notably allows uncensored outputs, so unlike the commercial video models out there, it can generate videos of anatomically realistic, nude humans.
OpenAI says it has no plans to release an API for Sora, its AI model that can generate reasonably realistic videos when provided with a text description or reference image. During an AMA with members of OpenAI’s dev team, Romain Huet, head of developer experience at OpenAI, said that a Sora API isn’t in the […]
Even before OpenAI's video generator Sora rolled out, bold-faced creatives were bracing for impact.
Costfoto/NurPhoto via Getty Images
Business Insider spoke to up-and-coming filmmakers and professors as OpenAI's Sora debuted.
AI video generation could open the door for indie filmmakers — and more blockbusters.
Fear of job losses looms, but one professor called AI text tools a bigger threat.
Up-and-coming filmmakers and professors at some of the nation's top film schools say the arrival of OpenAI video generators like Sora signals a democratization of the industry may be afoot, even though the tech is still limited.
Sora rolled out widely on Monday following a February pilot program. The tool generates short video clips — 20 seconds max — from users' text prompts. Sora can also modify existing clips.
For example, say a user wants to create a scene with green monsters in a thunderstorm. To do that, she'd type a prompt, and Sora would spit out a file.
Michaela Ternasky-Holland was one of the first directors to create and premiere a short film using Sora. It screened at Tribeca in 2024. She said she's excited about Sora's potential to cut filmmaking's development costs by creating things like sizzle reels, but she's aware of its limitations.
These things are giving you an illusion of control. And no matter how good the generations are, there's still someone behind them prompting it," she said. "Just because someone has a 4K camera, it doesn't make them a Steven Spielberg."
The tool can generate short video clips — 20 seconds max — from user-inputted text prompts.
Sora
Dana Polan, a professor of cinema studies at New York University's Tisch School of the Arts, said AI image generators aren't stoking the same fears as their text-based counterparts.
That's because many in Hollywood see the screenplay as "the first act of creativity," said Polan, who noted that other people in the filmmaking process, including cinematographers, are already seen as "adapters into images of words."
While he remains optimistic about AI in film, George Huang, a professor at the UCLA School of Theater, Film and Television — who has experimented with AI tools in his own moviemaking — concedes the technology has a bad rap in Hollywood, which has made countless movies on the topic.
"We think AI is now coming to destroy all of us, and that's a narrative that Hollywood created," he said. "It's embedded in our culture."
Sora's not quite ready for prime time — yet
Industry watchers told Business Insider that they don't foresee Sora or AI image generation appearing widely in finished films just yet given that the image quality still exists in something of an "uncanny valley."
Sora's pace of improvement has slowed down with later versions, Ternasky-Holland said. For example, it still struggles to put multiple characters in a scene no matter how many times it's prompted, she said.
Experts don't foresee Sora or AI image generation appearing widely in finished films just yet.
Sora
But Polan told BI the tech could come in handy for the previsualization process — or animated storyboards to check pacing and flow. Huang also said he could see it being used as a "pitch reel" for screenwriters.
That said, other AI startups like Runway have created tools already used across the industry to expedite editing, with clients that include "The Late Show with Stephen Colbert" and the effects team behind "Everything Everywhere All At Once."
Michael Gilkison, a Lexington, Kentucky-based filmmaker whose latest project, "The Finish Line," is on Amazon's Prime Video, said a free AI app helped create a scene where a car was crushed. "That would have cost a lot more 20 years ago," he said over email. Using AI technology could also create cheaper ways to film period pieces. But it also can negate the need to hire extras, which can deprive a film of its spirit.
"As a producer, I would use it to keep the cost down, but it is all about balance," Gilkison said.
Tahsis Fairley, a creative producing student at Chapman University, said via email he envisions using Sora to expedite storyboarding and illustrate ideas to his team.
"We will be able to test out new visual ideas without investing significant amounts of money," Fairley said.
That said, Huang doesn't believe we're far off from full implementation, saying AI could appear within completed films "by the end of the next year easily."
Cost savings could boost indies and blockbusters alike
The expenses associated with filmmaking can put a damper on artistic vision, Huang said. But students are generally receptive to new technology, Polan said.
ChatGPT Plus subscribers, who pay $20 a month, get up to 50 Sora generations a month that are five seconds maximum. ChatGPT Pro users, who pay $200 a month, get unlimited generations up to 20 seconds in length.
In slashing costs, Huang said platforms like Sora are bound to "almost democratize the filmmaking process, sort of lower those barriers to entry." In addition to more tools for indie filmmakers working in the margins, this could also mean more blockbusters produced at a relative discount by major studios, he said.
Fairley, for his part, sees AI as a "double-edged sword."
While he cheered its efficiency gains, he expressed concern about job losses across the industry — particularly in fields like animation, pointing to a Coca-Cola Chrismas ad created entirely with AI.
OpenAI did not immediately respond to a request for comment from Business Insider.
On Wednesday, a video from OpenAI's newly launched Sora AI video generator wentviral on social media, featuring a gymnast who sprouts extra limbs and briefly loses her head during what appears to be an Olympic-style floor routine.
As it turns out, the nonsensical synthesis errors in the video—what we like to call "jabberwockies"—hint at technical details about how AI video generators work and how they might get better in the future.
But before we dig into the details, let's take a look at the video.
OpenAI has never revealed exactly which data it used to train Sora, its video-generating AI. But from the looks of it, at least some of the data might’ve come from Twitch streams and walkthroughs of games. Sora launched on Monday, and I’ve been playing around with it for a bit (to the extent the capacity […]
On Monday, OpenAI released Sora, an AI video generator, in hopes of helping creators.
One such creative, Marques Brownlee, wants to know if his videos were used to train Sora.
"We don't know if it's too late to opt out," Brownlee said in his review of Sora.
On Monday, OpenAI released its Sora video generator to the public.
CEO Sam Altman showed off Sora's capabilities as part of "Shipmas," OpenAI's term for the 12 days of product launches and demos it's doing ahead of the holidays. The AI tool still has some quirks, but it can make videos of up to 20 seconds from a few words of instruction.
During the launch, Altman pitched Sora as an assistant for creators and said that helping them was important to OpenAI.
"There's a new kind of co-creative dynamic that we're seeing emerge between early testers that we think points to something interesting about AI creative tools and how people will use them," he said.
One such early tester was Marques Brownlee, whose tech reviews have garnered roughly 20 million subscribers on YouTube. One could say this is the kind of creator that OpenAI envisions "empowering," to borrow execs' term from the livestream.
But in his Sora review, posted on Monday, Brownlee didn't sugarcoat his skepticism, especially about how the model was trained. Were his own videos used without his knowledge?
This is a mystery, and a controversial one. OpenAI hasn't said much about how Sora is trained, though experts believe the startup downloaded vast quantities of YouTube videos as part of the model's training data. There's no legal precedent for this practice, but Brownlee said that to him, the lack of transparency was sketchy.
"We don't know if it's too late to opt out," Brownlee said.
In an email, an OpenAI spokesperson said Sora was trained using proprietary stock footage and videos available in the public domain, without commenting on Business Insider's specific questions.
In a blog post about some of Sora's technical development, OpenAI said the model was partly trained on "publicly available data, mostly collected from industry-standard machine learning datasets and web crawls."
Brownlee's big questions for OpenAI
Brownlee threw dozens of prompts at Sora, asking it to generate videos of pretty much anything he could think of, including a tech reviewer talking about a smartphone while sitting at a desk in front of two displays.
Sora's rendering was believable, down to the reviewer's gestures. But Brownlee noticed something curious: Sora added a small fake plant in the video that eerily matched Brownlee's own fake plant.
Sora included a fake plant in a video that was similar to Brownlee's own plant.
Marques Brownlee
The YouTuber showed all manner of "horrifying and inspiring" results from Sora, but this one seemed to stick with him. The plant looks generic, to be sure, but for Brownlee it's a reminder of the unknown behind these tools. The models don't create anything fundamentally novel; they're predicting frame after frame based on patterns they recognize from source material.
"Are my videos in that source material? Is this exact plant part of the source material? Is it just a coincidence?" Brownlee said. "I don't know." BI asked OpenAI about these specific questions, but the startup didn't address them.
Sora created a video of a tech reviewer with a phone.
Marques Brownlee
Brownlee discussed Sora's guardrails at some length. One feature, for example, can make videos from images that people upload, but it's pretty picky about weeding out copyrighted content.
A few commenters on Brownlee's video said they found it ironic that Sora was careful to steer clear of intellectual property — except for that of the people whose work was used to produce it.
"Somehow their rights dont matter one bit," one commenter said, "but uploading a Mickeymouse? You crook!"
In an email to BI, Brownlee said he was looking forward to seeing the conversation evolve.
Millions of people. All at once.
Overall, the YouTuber gave Sora a mixed review.
Outside of its inspiring features — it could help creatives find fresh starting points — Brownlee said he feared that Sora was a lot for humanity to digest right now.
Brownlee said the model did a good job of refusing to depict dangerous acts or use images of people without their consent. And though it's easy to crop out, it adds a watermark to the content it makes.
Sora's relative weaknesses might provide another layer of protection from misuse. In Brownlee's testing, the system struggled with object permanence and physics. Objects would pass through each other or disappear. Things might seem too slow, then suddenly too fast. Until the tech improves, at least, this could help people spot the difference between, for example, real and fake security footage.
But Brownlee said the videos would only get better.
"The craziest part of all of this is the fact that this tool, Sora, is going to be available to the public," he said, adding, "To millions of people. All at once."
He added, "It's still an extremely powerful tool that directly moves us further into the era of not being able to believe anything you see online."
Rohan Sahai, Sora's product lead, said a team of about five or six engineers built the video generator in months.
"Sora is a tool," Joey Flynn, Sora's product designer, said. "It allows you to be multiple places at once, try multiple ideas at once, try things that are entirely impossible before."
OpenAI showed off the new product and its various features during a livestream Monday with CEO Sam Altman.
A screenshot of Sora's "explore" page for browsing AI videos from the community.
OpenAI
Sora includes an "explore" page, a browsable feed of videos shared by the community. OpenAI also showed the various style presets available, such as pastel symmetry, film noir, and balloon world.
To customize videos further, there's also Storyboard, which lets users organize and edit sequences on a timeline. The feature helps pull together text prompts that Sora then builds into scenes.
The company showed off Sora's features, including Storyboard.
screenshot/OpenAI
In February, OpenAI made Sora available to a limited group of creators, including designers and filmmakers, to get feedback on the model.
The company said in a blog post at the time that the product "may struggle to simulate the physics of a complex scene" and may not understand cause and effect. It may also mix up left and right and struggle to depict events that happen over time, it added.
The tool has already made a strong impression on some in Hollywood. Tyler Perry previously put his plans for an $800 million studio expansion on hold after seeing Sora. The billionaire entertainer referred to Sora demonstrations as "shocking" and said AI would likely reduce the need for large sets and traveling to locations for shoots.
However, the tool's product designer said in the demonstration Monday that Sora wasn't going to create feature films at the click of a button. Flynn said the tool was more "an extension of the creator who's behind it."
"If you come into Sora with the expectation that you'll just be able to click a button and generate a feature film, I think you're coming in with the wrong expectation," Flynn added.
The team also briefly touched on safety issues. Sahai said during the presentation that OpenAI had a "big target" on its back and that the team wanted to prevent illegal activity while balancing creative expression with the new product.
"We're starting a little conservative, and so if our moderation doesn't quite get it right, just give us that feedback," Sahai said. "We'll be iterating."
OpenAI said Sora would roll out to the public in the US and many other countries on Monday. But Altman said it would be awhile before the tool became available in the UK and most of Europe.
ChatGPT Plus subscribers, who pay $20 monthly, can get up to 50 generations a month of AI videos that are five seconds long and have a 720p resolution. ChatGPT Pro users, who pay $200 a month, get unlimited generations in the slow-queue mode and 500 faster generations, Altman said in the demo. Pro users can generate up to 20-second-long videos that are 1080p resolution, without watermarks.
While nonpaying users can't create Sora videos, they can browse Sora's explore feed, Altman said.
The prominent YouTuber Marques Brownlee published what he described as the first-ever Sora review on Monday, telling his nearly 20 million subscribers that the results were both "horrifying and inspiring."
After a brief overview of Sora's strengths and weaknesses — the YouTuber said that it could make provocative videos of cosmic events in deep space and other abstractions but that it struggled with realistic depictions of physics in day-to-day life, like a man running with a football — Brownlee was frank about his concerns.
Millions of people can now use Sora for basically whatever they want. And while the program has decent guardrails, one can be circumvented, he said. The little watermark that Sora adds to the bottom-right corner of its videos can be cropped out, Brownlee said.
"And it's still an extremely powerful tool that directly moves us further into the era of not being able to believe anything you see online," he said, adding: "This is a lot for humanity to digest right now."
On Monday, OpenAI released Sora Turbo, a new version of its text-to-video generation model, making it available to ChatGPT Plus and Pro subscribers through a dedicated website. The model generates videos up to 20 seconds long at resolutions reaching 1080 p from a text or image prompt.
Open AI announced that Sora would be available today for ChatGPT Plus and Pro subscribers in the US and many parts of the world but is not yet available in Europe. As of early Monday afternoon, though, even existing Plus subscribers trying to use the tool are being presented with a message that "sign ups are temporarily unavailable" thanks to "heavy traffic."
Out of an abundance of caution, OpenAI is limiting Sora's ability to generate videos of people for the time being. At launch, uploads involving human subjects face restrictions while OpenAI refines its deepfake prevention systems. The platform also blocks content involving CSAM and sexual deepfakes. OpenAI says it maintains an active monitoring system and conducted testing to identify potential misuse scenarios before release.
Sora, OpenAI’s video generator, is launching Monday — at least for some users. YouTuber Marques Brownlee revealed the news in a video published to his channel this morning. Brownlee got early access to Sora, and gave his initial impressions in a 15-minute review. Sora lives on Sora.com, Brownlee said, the homepage for which shows a […]
Subscribers to OpenAI’s ChatGPT Pro and Plus plans will get access to Sora, OpenAI’s video generator, starting today — but only if they live in certain countries. During a livestream Monday afternoon, OpenAI revealed key information about Sora, including technical details. The version launching today, called Sora Turbo, can generate clips between 5 seconds and […]
The campaign included 12 days of product releases, demos, and new features.
On the final day, OpenAI previewed o3, its most advanced model yet.
OpenAI released new features and products ahead of the holidays, a campaign it called "Shipmas."
The company saved the most exciting news for the final day: a preview of o3, its most advanced model yet, which the company said could be available to the public as soon as the end of January.
Here's everything OpenAI has released so far for "Shipmas."
'Shipmas' Day 1
OpenAI started the promotion with a bang by releasing the full version of its latest reasoning model, o1.
OpenAI previewed o1 in September, describing it as a series of artificial-intelligence models "designed to spend more time thinking before they respond." Until now, only a limited version of these models was available to ChatGPT Plus and Team users.
Now, these users have access to the full capabilities of o1 models, which Altman said are faster, smarter, and easier to use than the preview. They're also multimodal, which means they can process images and texts jointly.
Max Schwarzer, a researcher at OpenAI, said the full version of o1 was updated based on user feedback from the preview version and said it's now more intelligent and accurate.
"We ran a pretty detailed suite of human evaluations for this model, and what we found was that it made major mistakes about 34% less often than o1 preview while thinking fully about 50% faster," he said.
Along with o1, OpenAI unveiled a new tier of ChatGPT called ChatGPT Pro. It's priced at $200 a month and includes unlimited access to the latest version of o1.
'Shipmas' Day 2
On Friday, OpenAI previewed an advancement that allows users to fine-tune o1 on their own datasets. Users can now leverage OpenAI's reinforcement-learning algorithms — which mimic the human trial-and-error learning process — to customize their own models.
The technology will be available to the public next year, allowing anyone from machine-learning engineers to genetic researchers to create domain-specific AI models. OpenAI has already partnered with the Reuters news agency to develop a legal assistant based on o1-mini. It has also partnered with the Lawrence Berkeley National Laboratory to develop computational methods for assessing rare genetic diseases.
'Shipmas' Day 3
The Explore page of OpenAI's Sora AI tool, which generates AI videos from text prompts.
screenshot/OpenAI
OpenAI announced on December 9 that its AI video generator Sora was launching to the public.
"We want our AIs to be able to understand video and generate video and I think it really will deeply change the way that we use computers," the CEO added.
Rohan Sahai, Sora's product lead, said a product team of about five or six engineers built the product in months.
The company showed off the new product and its various features, including the Explore page, which is a feed of videos shared by the community. It also showed various style presets available like pastel symmetry, film noir, and balloon world.
OpenAI showed off Sora's features, including Storyboard for further customizing AI videos.
screenshot/OpenAI
The team also gave a demo of Sora's Storyboard feature, which lets users organize and edit sequences on a timeline.
Sora is rolling out to the public in the US and many countries around the world. However, Altman said it will be "a while" before the tool rolls out in the UK and most of Europe.
ChatGPT Plus subscribers who pay $20 monthly can get up to 50 generations per month of AI videos that are 5 seconds long with a resolution of 720p. ChatGPT Pro users who pay $200 a month get unlimited generations in the slow queue mode and 500 faster generations, Altman said in the demo. Pro users can generate up to 20-second long videos that are 1080p resolution, without watermarks.
'Shipmas' Day 4
ChatGPT can provide more specific edit notes and run code using canvas.
OpenAI
OpenAI announced that it's bringing its collaborative canvas tool to all ChatGPT web users — with some updates.
The company demonstrated the tech in a holiday-themed walkthrough of some of its new capabilities. Canvas is an interface that turns ChatGPT into a writing or coding assistant on a project. OpenAI first launched it to ChatGPT Plus and Team users in October.
Starting Tuesday, canvas will be available to free web users who'll be able to select the tool from a drop-down of options on ChatGPT. The chatbot can load large bodies of text into the separate canvas window that appears next to the ongoing conversation thread.
Canvas can get even more intuitive in its responses with new updates, OpenAI said. To demonstrate, they uploaded an essay about Santa Claus's sleigh and asked ChatGPT to give its editing notes from the perspective of a physics professor.
For writers, it can craft entire bodies of text, make changes based on requests, and add emojis. Coders can run code in canvas to double-check that it's working properly.
'Shipmas' Day 5
All Apple users need to do is enable ChatGPT on their devices.
OpenAI 'Shipmas' Day 5
OpenAI talked about its integration with Apple for the iPhone, iPad, and macOS.
As part of the iOS 18.2 software update, Apple users can now access ChatGPT directly from Apple's operating systems without an OpenAI account. This new integration allows users to consult ChatGPT through Siri, especially for more complex questions.
They can also use ChatGPT to generate text through Apple's generative AI features, collectively called Apple Intelligence. The first of these features was introduced in October and included tools for proofreading and rewriting text, summarizing messages, and photo-editing features. They can also access ChatGPT through the camera control feature on the iPhone 16 to learn more about objects within the camera's view.
'Shipmas' Day 6
OpenAI launched video capabilities in ChatGPT's Advanced Voice Mode.
screenshot/OpenAI
OpenAI launched its highly anticipated video and screensharing capabilities in ChatGPT's Advanced Voice Mode.
The company originally teased the public with a glimpse of the chatbot's ability to "reason across" vision along with text and audio during OpenAI's Spring Update in May. However, Advanced Voice Mode didn't become available for users until September, and the video capabilities didn't start rolling out until December 12.
In the livestream demonstration on Thursday, ChatGPT helped guide an OpenAI employee through making pour-over coffee. The chatbot gave him feedback on his technique and answered questions about the process. During the Spring Update, OpenAI employees showed off the chatbot's ability to act as a math tutor and interpret emotions based on facial expressions.
Users can access the live video by selecting the Advanced Voice Mode icon in the ChatGPT app and then choosing the video button on the bottom-left of the screen. Users can share their screen with ChatGPT by hitting the drop-down menu and selecting "Share Screen."
'Shipmas' Day 7
OpenAI introduced Projects on Day 7 of "Shipmas"
screenshot/OpenAI
For "Shipmas" Day 7, OpenAI introduced Projects, a new way for users to "organize and customize" conversations within ChatGPT. The tool allows users to upload files and notes, store chats, and create custom instructions.
"This has been something we've been hearing from you for a while that you really want to see inside ChatGPT," OpenAI chief product officer Kevin Weil said. "So we can't wait to see what you do with it."
During the live stream demonstration, OpenAI employees showed a number of ways to use the feature, including organizing work presentations, home maintenance tasks, and programming.
The tool started to roll out to Plus, Pro, and Teams users on Friday. The company said in the demonstration it will roll out the tool to free users "as soon as possible."
'Shipmas' Day 8
OpenAI announced on Monday it is rolling out SearchGPT to all logged-in free users.
screenshot/OpenAI
OpenAI is rolling out ChatGPT search to all logged-in free users on ChatGPT, the company announced during its "Shipmas" livestream on Monday. The company previously launched the feature on October 31 to Plus and Team users, as well as waitlist users.
The new feature is also integrated into Advanced Voice Mode now. On the livestream, OpenAI employees showed off its ability to provide quick search results, search while users talk to ChatGPT, and act as a default search engine.
"What's really unique about ChatGPT search is the conversational nature," OpenAI's search product lead, Adam Fry, said.
The company also said it made Search faster and "better on mobile," including the addition of some new maps experiences. ChatGPT search feature is rolling out globally to all users with an account.
'Shipmas' Day 9
OpenAI announced tools geared towards developers.
screenshot/OpenAI
OpenAI launched tools geared toward developers on Tuesday.
It launched o1 out of preview in the API. OpenAI's o1 is its series of AI models designed to reason through complex tasks and solve more challenging problems. Developers have experimented with o1 preview since September to build agentic applications, customer support, and financial analysis, OpenAI employee Michelle Pokrass said.
The company also added some "core features" to o1 that it said developers had been asking for on the API, including function calling, structured outputs, vision inputs, and developer messages.
OpenAI also announced new SDKs and a new flow for getting an API key.
'Shipmas' Day 10
You can access ChatGPT through phone calls or WhatsApp.
screenshot/OpenAI
OpenAI is bringing ChatGPT to your phone through phone calls and WhatsApp messages.
"ChatGPT is great but if you don't have a consistent data connection, you might not have the best connection," OpenAI engineer Amadou Crookes said in the livestream. "And so if you have a phone line you can jump right into that experience."
You can add ChatGPT to your contacts or dial the number at 1-800-ChatGPT or 1-800-242-8478. The calling feature is only available for those living in the US. Those outside the US can message ChatGPT on WhatsApp.
OpenAI employees in the live stream demonstrated the calling feature on a range of devices including an iPhone, flip phone, and even a rotary phone. OpenAI product lead Kevin Weil said the feature came out of a hack-week project and was built just a few weeks ago.
'Shipmas' Day 11
Open AI's ChatGPT desktop program has new features.
screenshot/OpenAI
OpenAI focused on features for its desktop apps during Thursday's "Shipmas" reveal. Users can now see and automate their work on MacOS desktops with ChatGPT.
Additionally, users can click the "Works With Apps" button, which allows them to work with more coding apps, such as Textmate, BB Edit, PyCharm, and others. The desktop app will support Notion, Quip, and Apple Notes.
Also, the desktop app will have Advanced Voice Mode support.
The update became available for the MacOS desktop on Thursday. OpenAI CPO Kevin Weil said the Windows version is "coming soon."
'Shipmas' Day 12
Sam Altman and Mark Chen introduced the o3 and o3 mini models during a livestream on Friday.
screenshot/OpenAI
OpenAI finished its "12 days of Shipmas" campaign by introducing o3, the successor to the o1 model. The company first launched the o1 model in September and advertised its "enhanced reasoning capabilities."
The rollout includes the o3 and 03-mini models. Although "o2" should be the next model number, an OpenAI spokesperson told Bloomberg that it didn't use that name "out of respect' for the British telecommunications company.
Greg Kamradt of Arc Prize, which measures progress toward artificial general intelligence, appeared during the livestream and said o3 did notably better than o1 during tests by ARC-AGI.
OpenAI CEO Sam Altman said during the livestream that the models are available for public safety testing. He said OpenAI plans to launch the o3 mini model "around the end of January" and the o3 model "shortly after that."
In a post on X on Friday, Weil said the o3 model is a "massive step up from o1 on every one of our hardest benchmarks."
A group appears to have leaked access to Sora, OpenAI’s video generator, in protest of what it’s calling duplicity and “art washing” on OpenAI’s part. On Tuesday, the group published a project on the AI dev platform Hugging Face seemingly connected to OpenAI’s Sora API, which isn’t yet publicly available. Using their authentication tokens — […]
