Ad
Skip to content
Read full article about: OpenAI turns model compression into a talent hunt with its 16 MB "Parameter Golf" challenge

OpenAI challenges researchers to build the best language model in just 16 MB - and uses the competition to scout talent. In an open research competition called "Parameter Golf," OpenAI is asking developers to build the best possible language model under tight constraints: weights and training code combined must stay under 16 MB, and training can take no longer than ten minutes on eight H100 GPUs. Submissions are judged on compression performance against a fixed FineWeb dataset.

OpenAI is putting up one million dollars in computing credits through its partner Runpod. Top performers may get invited for job interviews - the company plans to hire a small group of junior researchers in June, including students and Olympiad winners. The GitHub repository includes baseline models, evaluation scripts, and a public leaderboard. Anyone 18 or older in supported countries can participate through April 30.

The competition for AI talent among big tech companies is more intense than ever. Meta has repeatedly poached top researchers from OpenAI, in some cases offering compensation packages reportedly worth up to 300 million dollars.

Pentagon plans to let AI companies train models on classified data

The US Department of War is working to set up secure environments where AI companies can train their models on classified data. Until now, models were only allowed to read classified data, not learn from it.

OpenAI ships GPT-5.4 mini and nano, faster and more capable but up to 4x pricier

OpenAI has released two new compact models—GPT-5.4 mini and nano—built for coding assistants, subagents, and computer control. GPT-5.4 mini nearly matches the full model’s performance, but both new models come with a steep price hike over their predecessors.

OpenAI's biggest problem may not be building AI but getting companies to actually use it beyond ChatGPT

OpenAI is pushing to get its AI into large companies faster through sales, partnerships, and capital. A 10-billion-dollar joint venture and a new deployment arm show where the real challenge lies: getting the technology integrated into actual company workflows.

OpenAI's own wellbeing advisors warned against erotic mode, called it a "sexy suicide coach"

OpenAI’s wellbeing advisory board reportedly voted unanimously against the company’s planned Adult Mode for ChatGPT. Internally, the company is struggling with an error-prone age detection system and unresolved safety issues.

Read full article about: OpenAI is reportedly planning to integrate its video AI Sora into ChatGPT

OpenAI is reportedly planning to fold its video AI Sora directly into ChatGPT. So far, Sora has only been available as a standalone mobile and web app. OpenAI originally pitched it as a viral hit and potential TikTok alternative, a strategy that seemed to work early on, partly thanks to massive copyright infringements.

That momentum didn't last. According to The Information, the app has slid from No. 1 to No. 165 in the Apple App Store since launching last fall. CEO Sam Altman reportedly admitted internally that hardly anyone was sharing videos publicly. Rolling Sora into ChatGPT might fix that: with around 920 million weekly active users, the move would naturally drive more video generation. The standalone app will stick around for now, The Information reports.

Google already offers video generation in Gemini, though with tight capacity limits and only for paying subscribers. OpenAI will likely go a similar route: the company is strapped for compute, burns through cash supporting the roughly 95 percent of free ChatGPT accounts, and video generation is especially resource-hungry.

Read full article about: Amazon gets court order blocking Perplexity's AI shopping agent

A federal court in San Francisco has granted Amazon an injunction against AI startup Perplexity, barring it from using its AI browser agent Comet to make purchases on Amazon.

Amazon sued Perplexity in November, accusing the startup of fraud because Comet didn't disclose when it was shopping on behalf of a real person and ignored Amazon's demands to stop. The case raises a growing legal question: how should courts handle AI agents taking on complex tasks like online shopping?

Judge Maxine Chesney ruled that Amazon presented strong evidence that Perplexity was accessing users' password-protected accounts with their permission but without Amazon's authorization. Perplexity must also delete any collected Amazon data and has one week to appeal.

There's an interesting wrinkle here: Amazon recently became a major investor in OpenAI, which also sees product research and online shopping as key AI chat features. So far, though, OpenAI reportedly hasn't cracked direct checkout in its chat interface. Amazon may be positioning itself to step in and own that piece of the puzzle.