Daily AI & Dev Digest: Claude 4 Elevates Coding, Google's Gemma 4 Unleashes On-Device AI Power
Catch up on the latest in AI: Anthropic launches Claude 4 with advanced coding and agent capabilities, while Google DeepMind expands its Gemma 4 open models for efficient on-device intelligence.
Welcome to your daily dose of AI and software development news! Today, we're diving into significant advancements from Anthropic with their new Claude 4 models, setting new benchmarks for coding and agentic workflows. Meanwhile, Google DeepMind continues to push the boundaries of accessible AI with the latest updates to their Gemma 4 open models, designed for maximum efficiency across a range of devices.
TL;DR
- Anthropic launched Claude Opus 4 and Claude Sonnet 4, boasting enhanced coding, advanced reasoning, and new API capabilities for powerful AI agents.
- Google DeepMind has significantly advanced its Gemma 4 open models, bringing frontier intelligence and multimodal capabilities to mobile, IoT, and personal computing devices.
Introducing Claude 4
Anthropic has unveiled its next generation of Claude models: Claude Opus 4 and Claude Sonnet 4, which are designed to establish new industry standards for coding, advanced reasoning, and AI agents. Claude Opus 4 is highlighted as the world's best coding model, showcasing sustained performance on complex and long-running tasks, as well as agent workflows. Claude Sonnet 4 represents a notable upgrade from its predecessor, Claude Sonnet 3.7, offering improved coding and reasoning while demonstrating more precise instruction following.
Accompanying these new models are several key features, including extended thinking with tool use (beta), enabling Claude to integrate tools like web search into its reasoning processes to enhance responses. The models also feature new model capabilities such as parallel tool use, precise instruction adherence, and significantly improved memory capabilities when given access to local files by developers. This allows Claude to extract and save key facts, maintaining continuity and building tacit knowledge over time. Furthermore, Claude Code is now generally available, supporting background tasks via GitHub Actions and offering native integrations with VS Code and JetBrains for seamless pair programming. Developers can also leverage four new API capabilities to build more powerful AI agents, including a code execution tool, MCP connector, Files API, and prompt caching for up to one hour.
Claude Opus 4 and Sonnet 4 are hybrid models offering both near-instant responses and extended thinking modes. These models are accessible through the Pro, Max, Team, and Enterprise Claude plans, with Sonnet 4 also available to free users. They are available on Anthropic's API, Amazon Bedrock, and Google Cloud's Vertex AI. Pricing for Opus 4 is $15/$75 per million tokens (input/output), and Sonnet 4 is priced at $3/$15, consistent with previous models.
Claude Opus 4 is positioned as the world’s best coding model, demonstrating sustained performance on complex, long-running tasks and agent workflows.
Gemma
Google DeepMind continues to advance its commitment to open AI with the latest iterations of Gemma, highlighted as their most capable open models. These models are engineered to help developers create AI applications that can run effectively across diverse platforms, from cloud servers to personal laptops and even mobile phones. A significant focus for Gemma 4 is on achieving maximum compute and memory efficiency, enabling a new level of intelligence for mobile and IoT devices. Simultaneously, it offers unprecedented intelligence-per-parameter, bringing frontier intelligence to personal computers.
The Gemma 4 family has seen several recent developments. In June 2026, DeepMind introduced DiffusionGemma, built upon the Gemma 4 family and cutting-edge Gemini Diffusion research, along with Gemma 4 QAT, which focuses on model compression for enhanced mobile and laptop efficiency. Also in June 2026, the Gemma 4 12B model was introduced as a unified, encoder-free multimodal model. These follow earlier advancements in May 2026 with the acceleration of Gemma 4 through faster inference using multi-token prediction drafters, and the initial introduction of Gemma 4 in April 2026, emphasizing its purpose-built design for advanced reasoning and agentic workflows. These continuous updates underscore Google DeepMind's drive to make powerful AI accessible and efficient across various computing environments.
Gemma 4 aims to bring frontier intelligence to personal computers and a new level of intelligence to mobile and IoT devices through maximum compute and memory efficiency.