Google Gemini ecosystem updates

Google expands Gemini ecosystem with new developer and on-device tools. - FunctionGemma: local command model - Runs offline on phones and IoT - Based on Gemma 3 270M variant - Converts natural language to executable actions - Gemini CLI Extension: Conductor (CDD) - Conductor scans repo and extracts context - Conductor builds context packages for Gemini - Gemini 3 Flash released for speed and low cost - Gemini 3 Flash: faster, cheaper, Frontier intelligence - Flash uses 30% fewer tokens vs 2.5 Pro - Flash pricing: $0.50 input / $3.00 output per million tokens - Benchmarks: GPQA Diamond 90.4% MMMU Pro 81.2% - Gemini App and Search use Flash as default - Google Labs CC: Gmail “Your Day Ahead” assistant - CC drafts emails, reads Gmail/Calendar/Drive - New “gem” workflow supports n8n-like custom apps - Gems include infograph-master prompt and rendering chain - Opal / N8N integration enables natural-language workflows
Key insights
FunctionGemma enables local NL-to-action on devices
Google announced FunctionGemma, a compact model version of Gemma 3 (270M) optimized for function-calling tasks that converts natural-language commands into executable actions. FunctionGemma is designed to run locally on smartphones, IoT devices, car systems and embedded boards (e.g., Jetson Nano, Raspberry Pi) without cloud connectivity, enabling commands such as playing a playlist and adjusting air conditioning to be executed on-device.
Conductor CLI brings context-driven development
Google released the Gemini CLI Extension named Conductor, promoting Context-Driven Development (CDD). Conductor reads a .conductor.yaml config, scans local code, auto-packages relevant files and docs into a contextual request, and sends that package to Gemini so model outputs are grounded in project structure. The tool addresses manual copying, context loss, and irrelevant information in AI-assisted coding workflows.
Gemini 3 Flash targets speed and low cost
Google launched Gemini 3 Flash positioned as 'Frontier Intelligence Built for Speed' with performance claimed to exceed Gemini 2.5 Pro in many tasks. Flash reportedly uses ~30% fewer tokens and runs ~3× faster on common tasks. Published pricing examples: input $0.50 and output $3.00 per million tokens. Benchmarks cited include GPQA Diamond 90.4% and MMMU Pro 81.2%. Google is deploying Flash across Gemini App, Search AI Mode, Google AI Studio, Vertex AI and Antigravity.
Google Labs CC acts as proactive Gmail assistant
Google Labs introduced CC, a Gemini-based Gmail assistant that sends a daily 'Your Day Ahead' briefing by aggregating Gmail, Calendar and Drive items, drafts replies, manages calendar entries, and adapts to user feedback. CC is experimental within Google Labs and currently available to select Google AI Premium subscribers in the US and Canada via a waitlist.
Gem workflow and gems enable n8n-like app building
Gemini's ecosystem expanded with a new 'gem' that supports visual workflows and app construction similar to n8n. Developers shared 'infograph master' gem prompts and workflows that chain Gemini planning, Nano Banana Pro image generation and Veo video rendering. Integration with Opal and N8N-style capabilities allows users to build and share interactive AI apps without consuming model quota for recipients.