Creator of Gemini. Deep integration with Google Workspace, Search, and Android. Offers both free Flash and paid Pro tiers.
Creator of Gemini. Deep integration with Google Workspace, Search, and Android. Offers both free Flash and paid Pro tiers.
All Google Plans(1)
Models from Google(2)
Google History Timeline(30)
Full Claude Platform generally available on Google Cloud Vertex AI
Anthropic expands multi-cloud availability with a new Fast Mode tier, Claude Opus 5.5 Fast at $8 per million input tokens.
SourceGemini conversational navigation comes to Apple CarPlay
Available to US users with global rollout underway. Personal Intelligence features announced for Android Auto the previous day were not included.
SourceOpen-sources TorchTPU and ships a TPU Agent Suite to ease PyTorch migration
TechCrunch reported Oct 20 that Google open-sourced TorchTPU, a PyTorch-on-TPU release where moving to TPUs typically needs only a configuration change with no rewrite, plus a TPU Agent Suite for finding and fixing migration errors, and updated LiteRT for phones, laptops and embedded devices.
SourceBegins rolling out Gemini Personal Intelligence to Android Auto
The Verge reported Oct 20 that Google began rolling out Gemini Personal Intelligence to Android Auto, drawing on Gmail, Calendar, Maps, Photos, YouTube and Search; users must enable the setting and personalized answers require a connected phone.
SourceAntigravity CLI v1.4.1 strengthens plugin validation and diff preview
Oct 17 Google Antigravity CLI v1.4.1 released; plugin validate extends to SKILL.md/agents/rules; inline diff preview with syntax highlighting; fixes Windows sandbox and MCP server status issues
SourceElorian Glimpse visual reasoning model
Oct 16 ex-Google DeepMind researcher Andrew Dai's Elorian releases Glimpse, $55M seed at $300M valuation
SourceGoogle DeepMind announces new Cambridge robotics lab
Google DeepMind on October 15 announced a new robotics lab in Cambridge, UK, led by former Boston Dynamics CTO Aaron Saunders, focused on training foundation models for robots in real environments and humanoid control; deep collaboration with UK universities and NVIDIA compute.
SourceYouTube launches Ask YouTube AI chatbot search on US mobile
YouTube on October 15 launched its Ask YouTube AI chatbot-style video search on US mobile, powered by a Gemini-based multimodal video retrieval system; YouTube Music simultaneously rolled out new labels for AI-generated music.
SourceGemini 4 Argon global GA with 1M token output limit
Google CEO Pichai and DeepMind's Hassabis jointly announced on the early morning of October 14 the global GA of Gemini 4 Argon, the first frontier model in the Gemini 4 family, via the Gemini API, AI Studio, Google Cloud's Gemini platform, Antigravity, the Gemini App, and AI Mode in Search. Output token limits are raised from 64,000 to 1,000,000 and the model joins the expanded Fairwind cyber defender program; 10,000 scientists get six months of free highest-tier Antigravity access.
SourceGoogle joins UK legislative hearing, supports AISI and intelligence agencies constraining frontier model deployment
At an October 13 UK parliamentary hearing on AI safety legislation, OpenAI and Anthropic said some AI safety rules should be legally binding; Google DeepMind and Meta stopped short of endorsing new legislation. Officials acknowledged the AI Security Institute can test advanced models only if companies provide access and cannot stop a release it considers dangerous; OpenAI proposed mandatory incident reporting as a starting point.
SourceGemini unified agent launches for enterprises
Accepts goals, plans work and connects enterprise systems, launching first for businesses
SourceEmbeddingGemma 2 released: 740M parameter Apache 2.0 on-device multimodal embedder
Maps text, code, images, video and audio into a single 768-dimensional embedding space; quantized full multimodal uses about 567MB on a Pixel 11 Pro.
SourceTightens Gemini app model access
From February 9 free users lose Pro and keep only Flash + Flash-Lite; AI Plus/Pro/Ultra get 2x/4x/5-20x compute quotas
SourceGoogle Workspace goes native on Markdown, with .md files in Docs and Drive
Pushed to more than 3 billion users globally; users can open, edit and comment on rendered Markdown inside Docs and Drive; default rollout completes within 15 days.
SourceGemini 4 Argon first opened to trusted cyber defenders in Fairwind program
Google DeepMind launched Gemini 4 Argon on September 30 with a 1M-token output cap and 77.9% on DeepSWE v1.1. Initial access is limited to trusted cyber defenders in the Fairwind program, followed by paid API customers and Google AI Ultra subscribers. Introductory API price is $2 input / $10 output per million tokens.
SourcePublishes SynthID Bio in Nature: first verifiable watermark for AI-designed proteins
On September 30 DeepMind published SynthID Bio in Nature, embedding hidden watermarks in AI-designed amino acid sequences and AlphaFold 3-predicted 3D structures; wet-lab validation on VEGF-A, SARS-CoV-2 RBD, and PD-L1 confirmed the marks do not impair binding.
SourceGoogle Project Suncatcher MVP to launch on SpaceX Falcon 9 on October 1
Suncatcher MVP launches with SpaceX Transporter-18 carrying 4 Trillium TPUs and about 1 kW of solar; each run is capped at about 15 minutes; Planet Labs supplies the satellite bus.
SourceGemini 4 enters post-training, targeted to launch far earlier than year end
9/24 new DeepMind chief Kavukcuoglu confirmed at The Information summit, target is to be back at the frontier within 6 months
SourceGoogle rolls out Gemini 3.8 Live with Live Avatar in Gemini Enterprise
Google on September 24 made Gemini 3.8 Live with Live Avatar generally available in Gemini Enterprise. The release layers 24 fps video generation on top of the voice model, with lip-sync across 97 languages and SynthID watermarks on every output.
SourceGoogle to launch Project Suncatcher MVP satellite on October 1
Google on September 24 announced that the first Project Suncatcher MVP satellite will launch on October 1 aboard a SpaceX Falcon 9 from Vandenberg, carrying four Trillium TPUs with about 1 kW of solar power and roughly 15-minute compute duty cycles.
SourceGoogle tests Gemini Call for Me for Pixel 11 owners
Google on September 24 began testing Call for Me for Pixel 11 owners in the U.S. with paid Gemini subscriptions. The AI places calls from the user's own number, navigates menus, waits on hold, and lets the user take over at any point.
SourceGoogle joins SAFA Frontier AI Standards Agency effort
Google, OpenAI, and Anthropic are advancing the SAFA Frontier AI Standards Agency for late 2026 or early 2027, operating independently without government oversight. They have approached former White House AI policy adviser Sriram Krishnan for the CEO role.
SourceGemini 3.8 Flash TTS and Flash-Lite TTS launch with character voices
On September 23 Google DeepMind shipped Gemini 3.8 Flash TTS and Flash-Lite TTS with directable scene dialogue and natural-language voice design. The work was led by Alan Cowen, ex-Hume AI. Both are live in Gemini API and AI Studio.
SourcePrivate AI Compute gains persistent server-side memory with on-device keys
On September 23 Google updated the Private AI Compute brief with a server-side memory layer; user data is sealed in cloud enclaves while keys stay on user devices, with Noise-Protocol sessions.
SourceWeatherNext 3 launches with 5 km resolution and hourly refreshes
On September 23 Google DeepMind and Google Research released WeatherNext 3, refreshing every hour at 5 km resolution with up to 50% better precipitation accuracy. It ships in Search, Maps, Gemini app, and Maps Platform Weather API.
SourceGoogle's Gemini reached three real companies in May red-team test
WSJ reported on September 19 that Gemini in Irregular's May CTF reached three real companies after a fictional target shared a name with a real business; Google was told in July and said no public disclosure was warranted.
SourceGemini 3.8 Live models launch
Gemini 3.8 Live + Extended Thinking, Extended Thinking tops speech quality at 82.6.
SourceDeepMind Institute launches with 30-day pre-release review proposal
The DeepMind Institute is co-directed by Shane Legg, James Manyika and Demis Hassabis. Hassabis proposes a US-led frontier AI standards body that reviews models up to 30 days before release.
SourceGoogle DeepMind launches Gemini 3.8 Live and Live Extended Thinking
Google on September 15 launched Gemini 3.8 Live and Extended Thinking; the Extended Thinking variant topped the Artificial Analysis Speech-to-Speech Index at 82.6, while the standard model lists at about $1.38/hour of audio input, roughly half of GPT-Live-1.
SourceReleased standalone Gemini desktop app for Windows
Supports Windows 10/11 with default Alt+Space global activation hotkey, customizable. Integrates prompt input, multi-model selection, voice input, conversation history, and image and video generation.
Source