Google has introduced a new ultra-light model called Gemma 3 270M ideal for smartphones and other on-device use cases. The open-source model is power-efficient and small enough to run locally in the absence of an Internet connection, as Google demonstrated in internal tests using a Pixel 9 Pro SoC. With just 270 million parameters, Gemma 3 270M is a fraction of the size of flagship LLMs, which typically have billions of parameters. While Google’s new model was not made for complex conversational use, it is “designed from the ground up for task-specific fine-tuning with strong instruction-following.” Continue reading Google Says New Gemma 3 Is Ideal for Mobile, Edge Devices

OpenAI is rolling a new foundation model, GPT-5, via API for developers and enterprise users in three branded sizes — gpt-5, gpt-5-mini and gpt-5-nano — “to give developers more flexibility to trade off performance, cost, and latency.” The company said Thursday that it is also making GPT‑5 available to all ChatGPT Plus, Pro, Team and Free tier users. Enterprise and Education tier users are promised access this week. While GPT‑5 in the API platform is the reasoning model that powers maximum performance in ChatGPT, “GPT‑5 in ChatGPT is a system of reasoning, non-reasoning, and router models,” OpenAI explains. Continue reading OpenAI Announces Launch of GPT-5 Model Across All Tiers

Anthropic has released Claude Opus 4.1, an upgrade to Opus 4 that reportedly improves on agentic tasks, computer coding and reasoning. Pricing has not increased from what customers were paying for Opus 4, and the company promises “substantially larger improvements to our models in the coming weeks.” The move comes as Anthropic nears a new funding round targeting $3 to $5 billion, which could place a valuation of up to $170 billion on the startup. Recurring revenue hit $5 billion as of late July, which could increase to $9 billion by the end of the year. Claude Opus 4.1 was released two days before OpenAI unleashed GPT-5, and performs comparably in coding benchmarks. Continue reading Anthropic Seeks to Raise $5 Billion, Debuts Claude Opus 4.1

Alibaba’s Qwen team has launched Qwen3-Coder, which it calls its “most agentic code model to date.” While it will be made available in multiple sizes, the most powerful variant — Qwen3-Coder-480B-A35B-Instruct — is being released first. The 480 billion parameter mixture-of-experts model has 35 billion active parameters supporting a context length of 256,000 tokens natively and 1 million tokens with extrapolation methods for “exceptional performance in both coding and agentic tasks,” explains the group, which claims the quasi-open source model has agentic coding, agentic browser use, and agentic tool use comparable to Anthropic’s proprietary Claude Sonnet 4. Continue reading Alibaba Is Rolling Out Its ‘Most Agentic Code Model to Date’

Multinational retail giant Walmart has created dozens of AI agents in the past months. Now the company is overhauling how the agents are organized in hopes of making them easier to use. The AI assistants will be sorted into four categories of “super agents” designed to interact with customers, vendors, retail employees and software engineers. The vendor category will serve both Walmart’s suppliers and third-party merchants who have digital storefronts at Walmart.com. According to the retailer, each group of super agents will draw on the capabilities of multiple behind-the-scenes agents and present them to users via a unified interface. Continue reading Walmart AI Super Agents Organized to Improve Ease-of-Use

AWS has released a new AI coding tool called Kiro in preview. This IDE for agent apps is described by some as a vibe coding platform. However, AWS says Kiro “goes way beyond,” getting prototypes into production systems with features such as specs and hooks. In fact, Kiro was designed specifically to reduce issues common to vibe coding, the process of creating software using AI agents reacting to natural language prompts. This makes it popular among non-coders, resulting in an often chaotic process that Kiro attempts to professionalize. Available for free during preview, Kiro supports most popular programming languages. Continue reading AWS Kiro Agentic AI Developer Tool Now Free During Preview

Software development platform Hugging Face is taking orders on Reachy Mini, a table-top robot that lets people use the latest AI models to develop, test, deploy, and share real-world AI applications from their desk. The tiny test subject is 11 inches at work and nine inches in sleep mode. Due to begin shipping later this summer, Reachy Mini comes in two configurations: a $299 Lite version that must be tethered to a computer running Mac or Linux OS (Windows coming soon) and a wireless $449 model that has a Raspberry Pi 5 single-board computer built-in. Continue reading Hugging Face Opens Preorders on New ‘Reachy Mini’ Robots

The European Union has published a General Purpose AI (GPAI) Code of Practice designed to help companies comply with the AI Act, which includes copyright protections and transparency requirements for advanced models. The Code of Practice bans training models on unauthorized materials and says companies must comply with copyright-holder requests to omit work from datasets. Developers are required to provide documentation describing the features of their AI models. The AI Act began taking effect in August 2024 and is being implemented gradually, with key transparency, governance and privacy provisions coming into force next month. Continue reading EU Releases AI Practices Code to Help with Legal Compliance

Apple has changed its European App Store policies in response to the Digital Markets Act, in hopes the move will help ward off a potential fine of up to $585 million for violating the 2022 law in the way it charges commissions from third-party developers selling apps through links in the App Store. The European Union threatens fines of up to $60 million per day for DMA violations. A European Commission spokesperson said the body is assessing whether Apple’s new terms bring the company into compliance. The Commission is requiring the company “to make a series of additional changes to the App Store,” explains Apple, adding that “we disagree with this outcome and plan to appeal.” Continue reading Apple Introduces More App Store Changes to Avoid EU Fines

Anthropic has updated its Claude AI chatbot with the ability to build, host and share AI-powered apps directly in Claude. Launching in beta, the new function builds upon the Artifacts feature Anthropic introduced last year, allowing users to see and interact with what they asked Claude to create. “Now developers can iterate faster on their AI apps without worrying about the complexity and cost of scaling for a growing audience,” according to Anthropic. The San Francisco-based AI startup adds that millions people have already used Claude to create more than 500 million artifacts — from productivity tools to educational games. Continue reading Anthropic’s Claude Chatbot Is Now a No-Code App Developer

In a move to attract more developers to Gemini, Google is releasing an open-source command line interface (CLI) that will be free for most developers. CLIs offer a means to communicate with operating systems, and can be used as alternatives or complementary to an integrated developer environment (IDE). Gemini CLI has agentic capabilities and can code and “so much more,” according to Google, which lists content generation, problem solving, deep research and task management among its uses. Gemini CLI provides “lightweight access to Gemini, giving you the most direct path from your prompt to our model.” Continue reading Google Bows Gemini Command Line Interface for Developers

Google DeepMind has released a new vision-language-action (VLA) model, Gemini Robotics On-Device, that can operate robots locally, controlling their movements without requiring an Internet connection or the cloud. Google says the software provides “general-purpose dexterity and fast task adaptation,” building on the March release of the first Gemini Robotics VLA model, which brought “Gemini 2.0’s multimodal reasoning and real-world understanding into the physical world.” Since the model operates independent of a data network, it’s useful for latency sensitive applications as well as low or no connectivity environments. Google is also releasing a Gemini Robotics SDK for developers. Continue reading Google Gemini Robotics On-Device Controls Robots Locally

Startup Zencoder (formerly For Good AI) has launched a cloud-based AI-powered E2E testing agent that simplifies the pipeline from initial code to production-ready applications. Now in public beta, Zentester tackles “verification,” which Zencoder founder and CEO Andrew Filev calls “the missing link” in scaling AI-created code from concept to market-ready app. That complicated process is often delayed by a bottleneck in final testing. Zentester is designed to take that late-stage verification process “from days to hours,” Filev says. Zentester has the typical agent superpowers — seeing and interacting as users do by clicking buttons, filling in forms and navigating workflows. Continue reading Zencoder Testing Agent Shaves Weeks Off App Development

Snapchat is offering experimental augmented reality and generative AI tools through its new Lens Studio iOS app and web tool. According to the company, “you can generate your own AI effects, add your dancing Bitmoji to the fun, and express yourself with Lenses that reflect your mood or an inside joke.” The app responds to text prompts, producing filters that can be published to Snapchat. Snap previously offered generative AI capabilities only to professional creators as part of its Lens Studio. The company has also launched an app that lets users read and reply to messages using their Apple Watch. For professional developers, Snap’s Lens Studio has added tools to build Bitmoji games. Continue reading GenAI Powers Snapchat Tools for Creating AR Lenses, Bitmoji

Nvidia is rolling out DGX Cloud Lepton, a platform that connects AI developers with GPU access available through various cloud providers. Nvidia calls it “a compute marketplace” that offers tens of thousands of GPUs through a global network that features Nvidia Cloud Partners (NCPs). Among them: CoreWeave, Crusoe, Firmus, Foxconn, GMI Cloud, Lambda, Nebius, Nscale, Softbank Corp. and Yotta Data Services — offering Nvidia Blackwell and other architecture GPUs. Developers can tap into GPU compute capacity in specific regions for both on-demand and long-term computing, Nvidia says, adding that it expects leading cloud computing providers to eventually sign on. Continue reading DGX Cloud Lepton: Nvidia’s New GPU Compute Marketplace