AllYouCanTech

Leaked GitHub Repository Uncovers OpenAI’s GPT-5 Models Prior to Official Reveal

OpenAI’s GPT-5 Models Surfaces on GitHub Prior to Official Announcement

In light of an upcoming major live stream from OpenAI scheduled for later today, GitHub has disclosed the much-awaited GPT-5 models from the company. Although the original blog post has been removed, users on Reddit managed to identify and preserve it before OpenAI introduces its next-gen GPTs.

Following several delays and the intention to consolidate the ChatGPT GPT models, OpenAI appears positioned to concentrate on four distinct options. The archived GitHub post (via The Verge) highlights significant advancements in reasoning, code quality, and the user experience provided by GPT-5.

The available models include:

– **gpt-5:** Tailored for logical reasoning and multi-step operations.
– **gpt-5-mini:** A more compact variant for budget-conscious uses.
– **gpt-5-nano:** Fine-tuned for speed, ideal for applications that demand low latency.
– **gpt-5-chat:** Crafted for sophisticated, natural, multimodal, and context-sensitive conversations suitable for enterprise solutions.

What stands out about the GPT-5 leak from OpenAI is its capability to manage intricate coding tasks with minimal prompts.

GitHub leak corroborates earlier reports on OpenAI’s GPT-5

Prior to this leak, The Information stated there were at least three new models on the horizon. At that point, the publication indicated that while multiple models would be accessible, OpenAI would consistently choose the optimal version based on user needs.

Nevertheless, OpenAI’s GPT-5 may not represent a monumental advancement akin to the transition from GPT-3 to GPT-4, owing to rising infrastructure expenses and limited data for training these new models. Nonetheless, users will soon witness the advantages of this new version, as it will be capable of interpreting text, images, and audio. The context window is expected to expand alongside enhancements in memory, allowing the AI to manage more information simultaneously.

OpenAI’s live stream is set to commence at 10 AM PT/1 PM ET. This announcement follows the company revealing its inaugural open-weight models, including versions that can be operated locally on a computer and even one for smartphone use, as well as mental health guardrails for ChatGPT.

BGR will provide all updates promptly after the conclusion of OpenAI’s livestream.

iOS 26 Proposes an Expanded Screen for Apple Watch Ultra 3 Along with Five Other Speculated Features

With iOS 26 beta 5 now released for developers, Apple has introduced various features and made several enhancements to the iPhone software experience. However, in addition to these new functionalities, the company has also disclosed an Apple Watch image with a resolution that does not correspond to any of its existing smartwatches.

The most plausible explanation is that this image refers to the eagerly awaited Apple Watch Ultra 3. According to MacRumors contributor Aaron Perris, the image showcases a resolution of 422 x 512, which might also be the resolution for the upcoming Apple Watch Ultra 3. For comparison, the current Ultra 2 offers a display resolution of 410 x 502.

While it’s improbable that Apple will alter the casing of the Apple Watch, there is a chance that the company will opt for thinner bezels, potentially resulting in a slightly larger display. Over the years, Apple has made minor tweaks to the display, enhancing the Apple Watch experience and allowing users to see more information at a glance.

Speculations suggest a significant year for Apple Watch Ultra enthusiasts.

If you are contemplating an upgrade to an Apple Watch Ultra, the third-generation model could be just what you need. Here are the expected enhancements for the 2025 edition:

– **5G support**: As per Bloomberg’s Mark Gurman, Apple intends to transition from Intel 4G chips to new MediaTek 5G modems. The watch is anticipated to incorporate 5G RedCap, a low-tier 5G service that promises to deliver substantially faster cellular speeds than currently available.
– **Blood pressure tracking**: The reporter also foresees the inclusion of blood pressure tracking capabilities in this model. Initially, this feature will only indicate whether one’s blood pressure falls within a normal range, akin to the sleep apnea detection functionality.
– **Enhanced display**: The Apple Watch Ultra 3 is expected to feature the same display technology as the Apple Watch Series 10, allowing the display to refresh every second, showing the seconds on the Watch Face without draining battery life.
– **Upgraded processor**: The Apple Watch Ultra 3 is projected to be equipped with a new S11 SoC, which could be enhanced to support some of the less demanding AI functionalities.
– **Remarkable battery life**: To uphold the impressive battery life of previous Apple Watch Ultra models, the company will need to continue improving both hardware and software features.

That summarizes what we know (and believe we know, based on rumors and reports) about the Apple Watch Ultra 3 so far. We will keep you informed as we uncover more details regarding this wearable.

Google Intends to Deploy Its New AI Coding Tool on In-House Initiatives

Google’s AI-enhanced coding assistant, Jules, has officially exited beta. This service, designed for direct integration with GitHub, was first revealed in December 2024 as a project under Google Labs. It transitioned to beta testing in May, following its unveiling at I/O 2025. Now, after several additional months with beta testers, Google has officially launched Jules for its paying customers.

The AI coding solution is available at various price tiers, including a complimentary plan limited to 15 individual daily tasks and three simultaneous tasks. Furthermore, there are paid versions that grant subscribers additional daily and concurrent tasks—though these are linked to Google’s AI Pro and Ultra plans, which cost between $124.99 and $199.99 monthly.

Jules is just one of Google’s coding options, as Gemini already supports coding as well, and the company has recently introduced Opal, an app designed for vibe-coding. What makes Jules particularly compelling is that Google representatives informed TechCrunch of their intention to utilize Jules in more of their own internal coding initiatives.

Asynchronous code development

In addition to Google’s confidence in its offering, what distinguishes Jules from other coding solutions in the current AI landscape is its ability to work asynchronously with developers. The tool utilizes your existing database, cloning it and deploying it on a Google Cloud machine. From this point, Jules can function smoothly on your code without impacting your database until you’re prepared to implement changes. Given the concerns surrounding AI coding tools—such as the AI application that inadvertently erased a user’s entire database—Jules is designed to enhance security for existing code by not operating directly on your live systems.

Naturally, as with any AI-driven technology, it is essential to have checks and safeguards in place to avoid losing any work. Additionally, with hackers discovering methods to exploit prompt injections to compromise AI, there are further considerations to be aware of.

Nonetheless, Google’s intention to adopt Jules as a fundamental tool for coding within its internal teams indicates that the tool has considerable value for developers eager to embrace AI-driven solutions. Now that it is publicly available as part of Google’s AI offerings, there is ample reason to explore it if you have an interest.

Google Unveils Functionality to Generate Illustrated AI Storybooks within the Gemini Application

open-source ChatGPT models while hinting at the significant GPT-5 enhancement. For instance, Google incorporated multiple AI agents into Google Cloud that assist app creators and businesses in developing applications more rapidly and intelligently. The organization also introduced an astonishing new AI world model named Genie 3, which enables interaction with AI-generated content. Lastly, Google presented a novel Storybook mode for Gemini, which allows users to craft illustrated stories for their children on any theme.

The prevailing AI models fall short of matching the creative capabilities of humans. They are capable of answering inquiries and addressing issues, but creativity is not their forte. When AI reaches that level, it will probably mark the beginning of artificial general intelligence (AGI), the forthcoming evolution of AI that will endow it with the creativity required to tackle any situation in a manner akin to a human.

Yet, the new Storybook mode in the Gemini app does not require the AI to exhibit particular creativity. It could be precisely what some parents seek to keep their kids occupied with a fresh bedtime story when they tire of their favorites. It can also serve as an excellent resource for generating new tales spontaneously when your child requests one, or to clarify more intricate concepts in a simpler way. Furthermore, Storybook not only crafts a distinctive digital book by adhering to a guideline but also supplies illustrated picture book visuals.