Tech giants release new AI models and tools for coding and wearables

Meta's efficient Muse models

Meta has introduced Muse Ember, which is its smallest and most efficient model to date. The company designed this specific model to handle complex reasoning and the repetitive tasks that AI agents perform throughout the day. These tasks include routing, extracting data, and making specific code fixes.

The model is built to provide high intelligence at a lower price point compared to competitors like Claude Haiku 5.5 and GPT-6 Luna. It features a 1M-token context window and can understand text, images, audio, and video simultaneously. This allows it to work across large sets of documents and various types of inputs.

In addition to the Ember model, Meta is expanding its Muse agent to wearable devices. Users can now download and install Muse on Apple Watch and Wear OS to perform tasks via voice or text. These capabilities include booking event tickets, auditing subscriptions, and tracking daily expenses.

OpenAI's multi-computer agents

OpenAI is updating its dot agents, which serve as always-on assistants within ChatGPT. Users can now connect a single dot agent to several of their own local computers. This allows the user to choose which specific machine the agent uses to complete a given task.

This update is specifically aimed at helping developers manage complex workflows. For example, a developer could ask the dot to test a change in Xcode on a second Mac while they continue to write code on their primary laptop. It can also be used to reproduce bugs on older machines or click through interfaces to find where they break.

By allowing the agent to interact with multiple environments, OpenAI is making it easier to handle isolated testing. This setup helps developers keep their main workspace clean while the AI handles the heavy lifting of testing and screenshot reporting on dedicated hardware.

Google Maps and AI Studio

Google is making it easier for developers to create location-based applications by integrating the Google Maps Platform into Google AI Studio. Developers can now build apps that monitor air quality, locate stores, or plan routes simply by describing their ideas in plain text. The system handles the configuration and API key management automatically.

The integration allows for a one-click setup where Gemini handles the technical heavy lifting. Developers can choose from various spatial tools, including 3D buildings, street view, and weather data. Google has already suggested several concepts, such as a pet grooming route optimizer and a city immersive tour.

This move aims to shorten the time between an idea and a functional app. By providing these tools directly in the AI Studio environment, Google is lowering the barrier for creating complex, map-powered software. It allows for faster prototyping of tools that require real-time location data.

Anthropic's heavy-duty models

Anthropic has released two new models, Claude Fable 5.2 and Claude Mythos 5.2, designed for advanced technical workflows. These models are built to handle demanding tasks in fields like robotics, mathematics, and scientific research. They are capable of performing complex technical tasks autonomously for long periods.

The company noted that Fable 5.2 performed at a level similar to Opus 5.5 at maximum effort but at roughly half the cost per task. However, because it includes strong safeguards for biology and cybersecurity, it is available to all users. Anthropic stated, "Fable has strong cybersecurity and biology safeguards to guard against misuse," as part of their release notes.

In contrast, Mythos 5.2 is designed for more sensitive work, such as vulnerability discovery and sensitive scientific research. This model has tighter access controls, and organizations must apply through specific verification programs to use it. Anthropic suggests that most users should stick to Claude Opus 5.5 for general coding before moving to these specialized models.