5 Things to Try with Gemini 3 Pro in Gemini CLI (developers.googleblog.com)

Google has integrated Gemini 3 Pro—their most capable multimodal, reasoning-focused model—directly into the Gemini CLI, letting developers use advanced AI reasoning, agentic coding, and tool orchestration from the terminal. Access is rolling out gradually; Google AI Ultra subscribers or anyone with a paid Gemini API key can upgrade to Gemini CLI 0.16.x (npm install -g @google/gemini-cli@latest), run /settings and toggle Preview features to enable Gemini 3 Pro. Rollout status is tracked on a GitHub discussion. The integration promises faster, more precise command execution, richer multimodal workflows, and smarter tool use across development tasks. Technically, Gemini 3 Pro can synthesize text, images, and code to scaffold entire runnable projects (not just single files), follow complex specs, and operate agentically—planning and executing multi-step work. Examples: generate a single-file Three.js photoreal 3D Golden Gate prototype (golden_gate_bridge.html) with CDN ES-module imports, Import Map for examples, GLSL water shader, ACES tone mapping, InstancedMesh for cars/birds, sliders for time/fog/traffic, and no build step; turn a hand-drawn UI sketch into HTML/CSS/JS by dragging the image into the terminal; translate natural-language requests into complex shell ops (e.g., automating git bisect); produce accurate code documentation; and orchestrate cross-service debugging (Cloud Run + Snyk) via CLI extensions. The net effect: the CLI becomes an intelligent partner that speeds prototyping, debugging, and routine engineering work.