Google Model Releases, Tracked

Google has launched Gemini Omni 1.1 Flash into General Availability, introducing 40-second video chain extensions, first-and-last frame controls, and a $0.03/s draft pricing tier.

Google's Gemini 3.8 Flash under high thinking effort generated a complete 1,708-line 3D open-world wilderness survival game in a single HTML file on the first attempt — featuring 2D Simplex noise terrain, dynamic biomes, unprompted Web Audio sound synthesis, a 6-slot hotbar, and full day/night lighting cycles.

Google released Gemini 3.8 Flash on September 2, 2026 — a 5.7-point jump over 3.7 Flash on autonomous software engineering at the same introductory price. A restricted Cyber variant, available only through the new Fairwind Program, targets government and critical infrastructure operators for agentic vulnerability remediation.

Google announced the expansion of Google Antigravity to enterprise customers on August 20, 2026. The agentic development platform is now integrated into eligible Gemini Enterprise app subscriptions, offering administrative spend controls and extensions for VS Code.

Google released Gemini 3.7 Flash on August 13, 2026 — a post-training update to its mid-tier workhorse model that doubles DeepSWE scores and nearly triples AutomationBench performance over Gemini 3.6 Flash, while launching at introductory pricing of $0.75/$3.75 per million tokens through the end of 2026.

Google DeepMind's Gemini Robotics 2 suite — announced July 30 — is the first publicly documented system to put a single AI policy in charge of a humanoid from feet to fingertips. The Embodied Reasoning model (ER 2) is available now in AI Studio. The full-body VLA and On-Device 2 are restricted to early-access partners, including Apptronik, whose Apollo 2 is the primary demo platform.

Google DeepMind released Gemini Robotics 2 on July 30 — a three-model suite for humanoid robot control. Gemini Robotics ER 2 is publicly available via AI Studio; the full-body VLA and On-Device 2 models are restricted to early-access hardware partners including Apptronik.

Google updated the Managed Agents feature in the Gemini Interactions API with two production-readiness upgrades: asynchronous background execution (jobs that survive dropped connections) and native remote MCP server support (no proxy middleware required). Gemini 3.6 Flash is now the default model powering both.

A hands-on Gemini 3.1 Pro benchmark: acting as an AI coding assistant, it generated a 1,421-line Three.js WebGL racing game in one shot — complete with a procedural Web Audio engine synth and canvas minimap that were never in the prompt. Five follow-up fix rounds to reach launch-ready. Full prompt, fix log, and playable demo inside.

A hands-on Gemini 3.1 Pro coding test: high thinking mode one-shot a good-looking 3D roguelike dungeon crawler in Three.js — full HUD, minimap, lighting, weapon view — but it took two more rounds of bug reports to get from 'looks great, can't move' to actually playable. Full prompt and every round inside.

Gemini 3.6 Flash generated a complete first-person 3D voxel survival shooter in a single HTML file via Antigravity. Low effort shipped with inverted mouse controls; high effort fixed it one-shot and added a working gun with recoil animation. Full prompt and breakdown inside.

Google released Gemini 3.6 Flash alongside 3.5 Flash-Lite and 3.5 Flash Cyber — matching its predecessor's intelligence score but cutting cost and output tokens, while the promised Gemini 3.5 Pro remains in partner testing.

Google released Gemini 3.5 Flash-Lite alongside 3.6 Flash — a budget model at $0.30/$2.50 per million tokens built for high-volume, low-reasoning work, with a full 1M-token context window.

Gemini 3.5 Flash Cyber finds, validates, and patches software vulnerabilities faster than Google's mainline Flash models — but it's restricted to governments and trusted partners, with no public API or pricing.