Grok 4.5 Enters Private Beta at Tesla and SpaceX, as Musk Claims AI Rivals Claude Opus

By Devanshi Kashyap | June 28, 2026

Elon Musk just announced that Grok 4.5, xAI’s newest artificial intelligence model, has entered private beta testing at Tesla and SpaceX. It is a big moment for xAI, since they are moving from training their models to actually using them inside Musk’s two tech giants. Right now, only internal teams can use Grok 4.5, but the news gives a peek at what xAI’s next generation AI can do and shows they are serious about matching up with the industry’s best. Musk says the model runs on a much larger, more advanced architecture and has gone through extra training to sharpen […]

Microsoft Launches MAI-Code-1-Flash on GitHub Copilot, Bringing a New In-House AI Coding Model

By Devanshi Kashyap | June 28, 2026

Microsoft has expanded the range of its AI models available in GitHub Copilot with the launch of MAI-Code-1-Flash, which is a new coding focused model developed by the company. The rollout introduces another option for developers using Copilot, highlighting Microsoft’s efforts to build its own AI technologies while enhancing the coding assistant’s growing multi model experience. Adding MAI-Code-1-Flash reflects Microsoft’s ongoing strategy of integrating proprietary AI models into its developer tools. Microsoft is not tying developers to one model, but instead, they are offering more than one way to get things done, and that kind of flexibility can actually change […]

Meta to its staff: “Be more efficient with AI tokens” on Using Gemini for its AI Projects

By Khwaish Manwani | June 28, 2026

Meta reports a cautious use of Google Gemini’s AI models after running into capacity limits. The staff were told to be more cautious with using the tokens, and the company continues to use internal tools for the project. The development focuses on a massive problem in the artificial intelligence industry. Even big companies can face architectural disadvantages when demand exceeds supply. It depicts how devoid compute capacity hinders workloads and pushes the organization to look for distinct methods. It also puts forward a question. If the access stays within constraints, will it still stay with Google or look for other […]

DeepSeek Launches DSpark Boosting – What It Is, How It Boosts Inference Upto 80%

By Khwaish Manwani | June 28, 2026

DeepSeek has launched DSpark, a new mechanism created to help large language models generate quick answers without affecting the model’s intelligence. The core idea is to predict the next token one by one instead of forcing it. DSpark has a semi-autoregressive graph that can predict multiple next tokens ahead and verify the ones that seem genuine. That reduces waste, computation, and back-and-forth oscillation, and helps attend to more users at once. DeepSeek says that this method can better inference speed by roughly 60 to 80 percent in real-time, making it crucial for throughput artificial intelligence systems. Why was DSpark Launched? […]

Claude’s Shift to Agentic Work Is Rewriting Anthropic’s Economic Index

By Khwaish Manwani | June 27, 2026

Anthropic is about to change how it calculates the economic impact of its artificial intelligence by swerving away from conversation logs towards more structured outputs that are a testament to real work. As Claude is extensively used for coding, research, and collaboration, Anthropic states that the conventional chat-based assessments no longer capture what the models are actually doing. Instead of assessing Claude as a message-centric application, the company is now assessing it as a system that produces useful metrics and outputs. This shift depicts a comprehensive view where Claude is becoming a window into how AI boosts economic activity, not […]

Google-Adds-AI-Guidance-to-Demand-Gen-Campaigns

Google Drops Gemini’s AI Insights into Demand Gen Ads

By Devanshi Kashyap | June 27, 2026

Google is rolling out new Gemini AI powered tools for its Demand Gen ad campaigns, making it easier for advertisers to set up and optimize their campaigns. The update gives marketers real time recommendations throughout the process: from picking creative assets to fine tuning their targeting. It is another sign that Google is focusing on artificial intelligence as a key part of its advertising platform. These updates aim to help businesses get past the usual learning curve and boost campaign results across Google’s visual platforms, like YouTube, Discover, Gmail, and the Google Display Network. Instead of leaving advertisers to rely […]

MirrorCode-The-New-AI-Benchmark-that-Tests-whether-Models-can-Recreate-Entire-Software-Programs

Epoch & METR Unveil 19-Day MirrorCode AI Test – How It Works

By Devanshi Kashyap | June 27, 2026

Artificial intelligence keeps getting better at writing code but mostly coding benchmarks focus on pretty small jobs like writing a single function, fixing a bug, or finishing off a code snippet. Those tests do a decent job at comparing different AI models, but they do not really tell us if an AI system can actually handle the bigger, long term task of putting together an entire software application from scratch. That is why Epoch and METR came up with MirrorCode, which is a new benchmark built to see how well AI performs on big, complex software engineering projects. MirrorCode is […]

the white house unrestricts mythos 5 for over 100 trusted organizations

Why the White House Gets to Gatekeep Mythos 5

By Devanshi Kashyap | June 27, 2026

The US government’s latest decision on Anthropic’s Claude Mythos 5 has started off another round of debate over who controls access to the world’s most advanced AI models. The Trump administration, after briefly restricting the model’s deployment because of national security concerns, just gave Anthropic the green light to restore access but only for around 100 selected and trusted US organizations. The restrictions stay the same for everyone else. The White House says it’s limiting access to Mythos to keep the AI out of the wrong hands but critics aren’t buying it. The core question is pretty simple: Should it […]

NVIDIA Aims To Prove Hardware and AI Models Must Align with NVFP4

By Khwaish Manwani | June 27, 2026

NVIDIA is clear that artificial intelligence performance is not about the algorithm nor big model. In a new deep dive on NVFP4 quantization, the company states that strategy, hardware, and pattern are one. The outcome is clear that a model is needed that understands decision laws and executes math. and information flows, and the model uses billions of parameters, the blend of software and hardware is a core aspect, not efficiency. NVIDIA introduces quantization to reduce the weight into bit-sized pieces for better memory and efficiency. It also has Nemotron 3 Ultra NVFP4, which runs on both of NVIDIA’s GPUs […]

In a recent talent war, OpenAI bagged Apple's Vision Pro Lead Paul Meade

OpenAI “Poaches” Apple’s Vision Pro, Smart Glasses Lead Paul Meade

By Abhijay Singh Rawat | June 27, 2026

In yet another talent coup, Apple is losing the executive who ran its most ambitious hardware bet. Paul Meade, the vice president in charge of the Vision Pro headset and Apple’s smart glasses efforts, is leaving for OpenAI, continuing a streak of high-profile defections to rivals in the AI and hardware sectors. He is set to leave Apple by next week and will then start at OpenAI’s hardware unit, working on the company’s upcoming family of AI-powered devices. As of yet, neither of the companies, nor Meade has commented publicly. But the move itself says something specific about where both […]

AI & Tech News