MentalHealthBench, Gemini TTS and Claude Marketplace land

OpenAI released MentalHealthBench on Sep 23 — an open benchmark co-built with more than 80 licensed psychologists and psychiatrists across 22 countries and 19 languages to score AI replies in realistic mental-health chats spanning non-acute, high-acuity, and emergency cases for adults, teens, caregivers, and clinicians; experts write weighted rubrics from −10 to +10, each conversation gets at least three reviewers, and GPT-5.6 Sol grades against the retained criteria across ten behavior dimensions, while OpenAI says ChatGPT is not a substitute for therapy or professional care (openai.com/index/introducing-mentalhealthbench/). The same day Google shipped Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS into Google AI Studio, the Gemini API, Gemini Notebook, and Google Vids (Flash-Lite on Gemini Enterprise “coming soon”) — Flash for generative character voices and line-by-line direction, Flash-Lite for high-volume cost-efficient speech — expanding past 30 presets to a 2,000+ production voice library and natural-language voice design across more than 100 languages, with 30-second voice replication gated by consent verification plus SynthID watermarking and C2PA credentials; Google cites #1 overall on Hume AI’s Voice Design Benchmark at 71.4 and #1/#2 on Hume’s Overall Quality Index, and lists voice-replication blocks for Illinois, Texas, the EEA, UK, Switzerland, and India (blog.google Gemini 3.8 text-to-speech). Anthropic opened the Claude Marketplace the same day as one place for more than 2,000 plugins and connectors (Atlassian, Google, Microsoft, Notion, Salesforce and others), Claude-powered agents and products from partners including CrowdStrike, Cursor, Harvey, Legora, Lovable, and Snowflake that teams can buy with a portion of committed Anthropic spend, and Claude Partner Network services from Accenture, Boston Consulting Group, and Deloitte (claude.com/blog/claude-marketplace).

Earlier coverage

Abbott TCEQ data-center permit halt; Muse human-concierge rollback; Snorkel $350M Series E; Unit 42 Continuous Frontier AI Defense; BC OpenAI Tumbler Ridge suit.

— Lark Pemberton