Mistral's Leanstral code verification agent outperforms Claude Sonnet at 15% of the cost

Google has deployed its second major Gemini for Home update this month, reducing response latency by 40% for common commands and substantially shortening voice responses for alarms, timers, and calendar events. The update adds five new alarm/timer capabilities including world event-based triggers and recurring alarms, while expanding Gemini Live translation to 30 languages and rolling out Home features to 19 additional countries.

OpenAI has released GPT-5.4 mini, the lightweight variant of its agentic coding model GPT-5.4, in GitHub Copilot. The model represents OpenAI's highest-performing mini offering to date for code generation and completion tasks.

Google announced Tuesday that all US users, including free-tier subscribers, now have access to Personal Intelligence in the Gemini app, Chrome, and AI Mode in Search. The feature, previously limited to paid AI Pro and AI Ultra subscribers, connects Gmail, Google Photos, YouTube, and other Google services to automatically personalize Gemini's responses without manual prompt engineering.

Microsoft has launched Copilot Health, an AI-powered tool designed to integrate fitness data from 50+ devices, medical records from 50,000+ US healthcare providers, and health history into unified summaries. The company plans to charge for access via subscription after a free trial period, though pricing remains undisclosed.

Perplexity made Computer for Enterprise generally available to enterprise customers on March 12, claiming an internal study of 16,000+ queries showed $1.6 million in labor cost savings and 3.2 years of equivalent work completed in four weeks. The service integrates with Gmail, Outlook, GitHub, Linear, Slack, Notion, Snowflake, Databricks, and Salesforce, orchestrating tasks across 20 frontier models with agentic internet access.

OpenAI and Anthropic are shifting their flagship AI chatbots from text-only interfaces to interactive visual learning tools. ChatGPT now offers dynamic explanations for over 70 math and science concepts with adjustable variables, while Claude generates custom charts and diagrams inline within conversations.

Anthropic has launched Code Review, a new service that deploys multiple specialized agents to scan pull requests for bugs, security vulnerabilities, and regressions. According to Anthropic, reviews average $15-25 per PR and take approximately 20 minutes to complete, scaling with code complexity.

Alibaba has consolidated its AI operations into a new business unit called "Alibaba Token Hub" (ATH), reporting directly to CEO Eddie Wu. The restructuring merges the Qwen research team, consumer apps, DingTalk communication platform, and Quark-branded devices to accelerate collaboration and monetization across the company.

OpenAI confirmed its forthcoming "adult mode" will permit text-based erotic conversations in ChatGPT but explicitly block generation of pornographic images, audio, and video. The feature, first announced by CEO Sam Altman in October 2024, has been delayed multiple times—most recently in March 2025—as the company grapples with safety concerns including a 12% error rate in age verification systems.

Mistral AI has released Mistral Small 4, the latest iteration of its small-scale language model. The company claims improvements in reasoning and coding capabilities, though specific benchmark scores and pricing details have not been publicly disclosed.

OpenAI Codex subagents reached general availability after weeks of preview, enabling developers to define custom agents as TOML files and parallelize task execution. The feature mirrors Claude Code's implementation with default subagents for exploration, worker, and default operations.

Anthropic is temporarily doubling Claude's usage limits outside peak hours through March 27, 2026. The promotion applies to Free, Pro, Max, and Team plans, with non-peak hour limits increasing from 5 to 10 hours. Peak hours (weekdays 8am-2pm) remain unchanged.