US AI Company Product Matrix · Updated 11-20

Google

Google's play is"Embedded distribution": no standalone-app user acquisition — AI is built into Search, Android, and Workspace, products users already use. Acquisition cost approaches zero; the price is an innovation pace dragged down by big-company process.

Genie 4 lands on the Gemini API and Google AI Studio on Nov 19 (Interactions API, model ID models/genie-4, paid-tier API key required): generates interactive worlds at 720p/24fps with synchronized native audio, up to 120 seconds per session, ~$0.10/second (official Google blog, 11-19)

Key numbersAt a glance

1 billion+
AI Mode MAU (announced at I/O 2026)
1 billion+
Gemini App MAU
$190B
2026 capex guidance (Q2 earnings)
100 items
Total launches across two days of I/O 2026
$0.75 / $3.75
Gemini 3.8 Flash API per million tokens (input/output) · official standard pricing (through 2026-12-31; $1.50/$7.50 from 2027-01-01)
$2 / $10
Gemini 4 Argon limited-time API pricing per million tokens (input/output, through 11-30; $4/$20 after; 95% discount on cached input)
59
Gemini 4 Carbon's Artificial Analysis Intelligence Index v5.0 score (above Argon's 50; broadly available from 11-10, intro pricing $2/$10, $4/$20 after 11-30)
$1.50 / $7.50
Gemini 4 Hydrogen API per million tokens (input/output); 1M context; 51 on the Artificial Analysis Intelligence Index high tier (above Argon's 50), 54 on xhigh
$3 / $15
Gemini 4 Neon API per million tokens (input/output); 1M context; $0.15 cached input (95% discount); 61 on the Artificial Analysis Intelligence Index xhigh tier, ranked #3 overall, #1 on the coding agent index at 58.6; tiered comparison — low 44 / medium 55 / high 60 / xhigh 61 — Carbon leads on low, ties Neon on high tiers; Google's own model card says Neon leads Claude Opus 5.5 and GPT-6.2 Sol on 16 of 19 benchmarks, GraphWalks 95.2 (Opus 66.8, Sol 80.1)
1,000 questions
Max questions Gemini Decision 1 evaluates in a single call (same multimodal input; 150–250ms end-to-end latency; $0.10 per million input tokens, output free)
~2 minutes
Genie 4 scene coherence duration (twice Genie 3)
~$0.10/second
Genie 4 API pricing (interactive worlds at 720p/24fps with synchronized native audio, up to 120 seconds per session; live on the Gemini API and Google AI Studio from Nov 19)
$19B
Google's credit backing of Anthropic's TeraWulf 20-year compute lease (401MW, Kentucky; Google lends credit, not cash)
6 countries
Gemini Tasks free rollout — first countries (US, Canada, Australia, India, Japan, South Korea), built on Gemini 4 Hydrogen, runs multi-step tasks in the background; US users first get call-making (calling businesses for quotes and bookings), Google Pay checkout, and password-manager login; supports Gmail, Calendar, Drive and 60+ third-party apps

Product layersAll entries

1 · Model layer

Flagship: Gemini 4 Neon arrives
Gemini 4 NeonLaunched
Released Nov 18 (official Google blog): positioned between Carbon (hardest long-horizon tasks) and Hydrogen (fast everyday work); priced at $3/$15 per million tokens (input/output), $0.15 cached input (95% discount); 1M context; 61 on the Artificial Analysis Intelligence Index xhigh tier, ranked #3 overall, #1 on the coding agent index at 58.6; live the same day on the Gemini API (AI Studio), Gemini for Business, Antigravity (Pro/Ultra) and the Gemini App (Pro/Ultra), replacing Gemini 3.1 Pro and Argon; a cybersecurity edition follows within days to trusted defenders via the Fairwind Program; from Dec 1, Carbon/Argon intro pricing ends and rises to $4/$20, leaving Neon 25% cheaper than Carbon; follow-up benchmarks, Nov 19: Artificial Analysis tiered comparison puts Neon at low 44 / medium 55 / high 60 / xhigh 61 on the Intelligence Index, Carbon leading on low and tying Neon on high tiers; Google's own model card says Neon leads Claude Opus 5.5 and GPT-6.2 Sol on 16 of 19 benchmarks, GraphWalks 95.2 (Opus 66.8, Sol 80.1)
Announced on the official Google blog, 11-18
Gemini 4 ArgonLaunched
Globally GA on Oct 13, via six entry points — Gemini API, AI Studio, Antigravity, Gemini App, Search AI Mode, Gemini Enterprise; limited-time pricing $2 / $10 per million tokens (input/output, rising to $4 / $20 after Nov 30), 95% discount on cached input; unified 1,048,576-token context (default synchronous output cap 65,536 tokens); replaced by Gemini 4 Neon from Nov 18 (official Google blog)
Google claims #1 on the Vals Index; 100K+ employees use Antigravity monthly, agentic code commits +35%, rollbacks -30%; 6 months of top-tier Antigravity gifted to 10,000 scientists; Fairwind program opening to hundreds of new partners; embodied/robotics reasoning announced 10-16 (DeepMind blog): 87% on pure-reasoning ERQA (official self-test, N=400); model ID gemini-4-argon; opened to Google AI Ultra subscribers in AI Studio / Antigravity, with Interactions API and the google-genai SDK supporting it on day one in the Gemini API paid tier; demos include Boston Dynamics Spot quadruped, 6-DoF robotic arm, Franka FR3 dual-arm + MuJoCo simulation rehearsal; GitHub Copilot retires it Dec 14 (replaced by Gemini 4 Carbon; GitHub announcement, Nov 13)
Gemini 4 CarbonLaunched
Released Oct 28: frontier model for cybersecurity — vulnerability discovery, exploitation, and patching — plus complex long-horizon engineering and coding tasks; open broadly from Nov 10 to paid developers, enterprises, and AI Ultra subscribers (previously only select cybersecurity defenders in the Fairwind Program); intro pricing $2 / $10 per million tokens (input/output), rising to $4 / $20 after Nov 30; 1M context; 59 on the Artificial Analysis Intelligence Index v5.0 (above Argon's 50); also launched a new multimodal knowledge-agent benchmark, MKA (Carbon ranks second on quality at about one-sixth the cost of the top-scoring model), plus a /swarm multi-agent collaboration preview for Antigravity
Reuters, 10-28: more than 2x Argon's success rate on secure code generation in internal benchmarks; Artificial Analysis Index v5.0 scored it 59 (high-reasoning tier) — no uncertainty estimate; broad opening and pricing per Unite.AI and the Google blog, 11-10; 60.4% on CursorBench — first (Cursor's figures); Perplexity has rolled Carbon out to Pro/Max users; the AI Mode help page added a country list on Nov 11 — 122 countries/regions, excluding the 27 EU member states plus Norway, Iceland, Russia, China, Turkey, and Hong Kong; Carbon is limited to Google AI Ultra ($99.99/month and up)
Gemini Decision 1Launched
Released Nov 11: powers the Decisions API — evaluates up to 1,000 yes/no, multiple-choice, or rating questions on the same multimodal input (text / image / audio / video / PDF, up to 256K tokens) in a single call, returning probabilities and confidence from one forward pass, with 150–250ms end-to-end latency; pricing $0.10 per million input tokens, output $0; available now via the Gemini API (AI Studio, v1beta), with Google Cloud support to follow
Announced on the official Google blog, 11-11
AlphaGenome 2Launched
Released Nov 11: predicts the effects of DNA variants by cell type, trained on a 100M+ single-cell atlas (68 tissues); on the TenK10K test it identifies pathogenic expression variants 30% of the time (previous generation: 16%) with 43% accuracy on effect direction (previous generation: 23%); available now for non-commercial research via the AlphaGenome API, with commercial research in private preview; research use only, not clinically validated
Announced on the official DeepMind blog, 11-11
Gemini Robotics 2.5Launched
Released Nov 11: the strongest vision-language-action (VLA) model yet — success rates up 30%–100% on 200+ tasks, 20%–30% faster; the Apptronik Apollo 2 humanoid hit 96% success sorting auto parts on a live Mercedes production line; the ASIMOV-Jailbreak safety benchmark was open-sourced alongside, with Gemini 4 Argon resisting jailbreaks 99.5% of the time (baseline 82%); partner-only (Apptronik, Boston Dynamics, Agile Robots, and others), no public API
Announced on the official DeepMind blog, 11-11
Gemini 4 HydrogenLaunched
Released Nov 12: built for fast everyday coding and agentic workflows; 1M context; API pricing $1.50/$7.50 per million tokens (input/output); 51 on the Artificial Analysis Intelligence Index high tier (above Argon's 50), 54 on xhigh; rolling out now across the Gemini API, AI Studio, the Gemini app, Search, and Gemini for business; Antigravity opens it to AI Pro/Ultra subscribers, with Ultra-only Orchestrator (dynamically switching between Carbon and Hydrogen); replaces the Gemini 3.8 Flash tier; Google claims 57% on Terminal-Bench 4.1 (Gemini 3.8 Flash: 18.2%) and 2.5× the task speed of 3.8 Flash
Announced on the official Google blog, 11-12; cloud docs flag it as "short-term GA, at least 45 days' notice before retirement", with no RAG Engine, fine-tuning, or Live API support; available on GitHub Copilot from Nov 13; Copilot retires Gemini 3.8 Flash and Gemini 4 Argon on Dec 14, replaced by Hydrogen and Carbon respectively (GitHub announcement, Nov 13)
Gemini 3.8 FlashLaunched
Became the default fast model on Sept 2; the default base for the Antigravity API
90.8% on Terminal-Bench 2.1 · API $0.75 / $3.75 per million tokens (official pricing through 2026-12-31); GitHub Copilot retires it Dec 14 (replaced by Gemini 4 Hydrogen; GitHub announcement, Nov 13)
Gemini 3.8 LiveLaunched
Real-time voice model — reason while talking, call tools asynchronously in the background; also available in an Extended Thinking deep edition
Tops the Artificial Analysis voice leaderboard at 82.6 · 97 languages · released Sept 16
Deep ThinkLaunched
Deep-reasoning mode for complex multi-step tasks
Highest usage reserved for AI Ultra
Gemini 3.5 ProLaunched
Google confirmed it is skipping a generation: the 3.5 Pro planned for June will not ship; resources redirected to Gemini 4 (announced Sept 30)
Verified negative fact; the earlier "in partner testing" claim is void
Gemini OmniLaunched
Multimodal video-generation model, announced at I/O 2026 via the official Google and DeepMind X accounts
Omni Flash rolled out to paid users, Flow, and YouTube Shorts
Gemini Omni 1.2 FlashLaunched
Released Nov 5: full GA in the Gemini API and enterprise Gemini; native video generation/editing up to 15 seconds (10 seconds in the previous gen), in 1-second steps from 3–15 seconds, multi-turn conversational editing, stronger consistency; 720p at $0.07/second, 30% cheaper than 1.1; 1.1 requests auto-redirect to 1.2 from Dec 3
Announced on the official Google blog, 11-05

2 · Developer layer

Agentification is the main thread
Antigravity 2.0Launched
AI-first IDE released end-2025; upgraded at I/O 2026 from a VS Code plugin to a standalone desktop app
Native AGENTS.md support · official site antigravity.google
Antigravity CLI agyLaunched
Terminal agent platform rewritten in Go; fully replaces the old Gemini CLI; multi-agent parallel orchestration is the headline
Old CLI discontinued for consumers on 2026-06-18
Interactions API · Managed AgentIn Preview
One API call spins up a cloud Linux sandbox where the agent writes code, runs tests, and calls tools
Google foots the sandbox compute bill during the preview · launched Sept 15
Agent Compute EngineIn Preview
Announced Oct 22: one hardware-isolated microVM per agent, zero-credential sandbox, customer preview in the coming months
Announced by Google Cloud, reported by TechTimes
AI StudioLaunched
Free developer playground wired directly into the Gemini API (incl. 3.8 Live)
The default entry point for indie developers
Nano Banana 2.1 (gemini-nano-banana-2.1)Launched
GA on Oct 6: efficient image generation + multi-turn conversational editing, with improved quality, prompt adherence, multi-turn character consistency, and text rendering; aspect ratios 1:4 to 8:1, 1K / 2K / 4K support
Official Gemini API changelog (10-06, parsed by Releasebot); gemini-3.1-flash-image (Nano Banana 2) announced deprecated at the same time, shut down 2026-10-29
EmbeddingGemma 2Launched
Released Oct 6: 740M-parameter open-weights (Apache 2.0) multimodal embedding model; text/code/image/video/audio in a unified 768-dim embedding; runs on-device (~191MB text-only / ~567MB full-modal on Pixel 11 Pro)
AI Edge Gallery app adds on-device demos "Instant Media Search" and "Video Moments Finder"; reported by unite.ai and verdict.co.uk
Gemini Enterprise Agent Platform / Gemini EnterpriseLaunched
Enterprise API hosting and private deployments (formerly Vertex AI, renamed at Google Cloud Next on 2026-04-22); Live models in enterprise private beta
Serves large customers' compliance and compute needs

3 · Creative tools layer

Video is the big bet
Flow + Veo 3.1Launched
AI video generation and editing workbench, built on the Veo 3.1 video model
1,000 AI credits/month on Pro; 10,000 on the $100 Ultra tier, 25,000 on the $200 tier
Gemini NotebookLaunched
Formerly NotebookLM, officially renamed July 16; research + knowledge-base scenarios, with cinematic Video Overviews; became a Connected App on Oct 23 — manage and query notebooks in natural language mid-conversation, on by default globally (except EEA)
Folded into the Gemini App's "Notebooks" tab · ~1M-token context
JulesLaunched
Async AI coding assistant; handles coding tasks autonomously in the background
Limited usage on Pro / higher usage on Ultra
Nano Banana ProLaunched
Image-generation model; works with Veo 3.1 and Gemini 3 for video summaries
One link in the multimodal creative chain
Project GenieLaunched
Interactive world-building, officially announced by DeepMind at I/O 2026
Uses Street View data to generate interactive worlds · available globally on $200 Ultra
Genie 4Launched
Live on the Gemini API and Google AI Studio from Nov 19 (Interactions API, model ID models/genie-4, paid-tier API key required): generates interactive worlds at 720p/24fps with synchronized native audio, up to 120 seconds per session, ~$0.10/second; the Oct 27 research preview was the first real-time world model that natively generates sound — ambient audio, music, and character dialogue; change weather, move the camera, add characters mid-run in natural language; scenes stay coherent for ~2 minutes (twice Genie 3); all generated video and audio carry SynthID watermarks
Gemini 3.8 Flash decomposes tasks in the loop and verifies frame by frame; Gemini Omni anchors look and behavior; first to invited trusted testers, then Google AI Ultra subscribers and early Gemini API partners; aimed at film pre-visualization, game prototyping, and robot simulation; long-horizon drift, object permanence, and legible text remain issues; DeepMind announcement covered by multiple tech outlets
Gemini business toolsLaunched
Launched Oct 21: business landing pages, Nano Banana 2.1 brand-marketing visuals, and spreadsheet/invoice financial analysis for SMBs
Integrations with Adobe / Canva / Dropbox / Shopify / Wix / Zoho / Xero; rolling out globally; reported by The Hindu BusinessLine

4 · Embedded distribution layer

Google's ace: zero acquisition cost
K-12 school default-off policyLaunched
Announced Nov 3: Gemini, Gemini Notebook, Gemini in Classroom, and Search will be off by default for K-12 student accounts and must be enabled one by one by admins, effective early 2027; a new Search version pilots in select US schools within November, and Guided Learning launched globally in Classroom on 11-03; but because Search and AI are deeply integrated, schools that decline AI tools will lose Search as well — Google offers no Search version without AI Overviews / AI Mode; Google clarified on X on 11-03 that schools can enable Gemini, NotebookLM, and Search individually, in any combination
TechCrunch, MediaNama, Gizmodo, 11-04 (citing Google education lead Chris Phillips' 11-03 blog and X clarification); Senators Sanders and Blumenthal and former FTC chair Lina Khan attacked it as coercing schools into adopting AI
Search AI ModeLaunched
AI Q&A mode inside Search — no need to "download an AI app"
1B+ MAU (announced at I/O 2026)
Gemini AppLaunched
Primary chat entry point, web + mobile; from Oct 23, apps built in Gemini can be published to Google Sites in one click, staying interactive with no code copying
MAU 400M → 1B+ (announced by Pichai on X, 2026-08-11 — Google's fastest ever); one-click Sites publishing per official Google Workspace Updates
Gemini on Maps for Apple CarPlayLaunched
Rolling out in the US and globally from Oct 22: a Gemini icon in the Google Maps interface, callable directly from navigation
Reported by 9to5google and techbriefly
Personal Intelligence × Google Maps extensionLaunched
Oct 22: Personal Intelligence connects to Maps search and saved places, combining with Calendar for personalized recommendations
Launch fact; reported by 9to5google
Personal Intelligence × Google WalletLaunched
Personal Intelligence adds a Google Wallet app: US users can have Gemini analyze spending, track subscription price hikes, and generate cash-flow charts, and link bank accounts and credit cards via Plaid; read-only — no money transfers or investing; rolling out to all US Gemini users, any tier
Neowin, Nov 14
ChromeOS 156 × GeminiLaunched
Released Oct 24: Gemini media understanding on Chromebook Plus, Gemini features in Chrome files/quizzes, talk-to-type dictation, and video summaries
Reported by ghacks.net
Android system-level integrationLaunched
Reaches billions of devices via system updates — the deepest moat
Users gain AI capabilities without noticing
The full Workspace suiteLaunched
Gemini embedded across Gmail / Docs / Sheets / Drive / Meet
Enterprise switching costs are extremely high; the strongest lock-in
Gemini enterprise general agentLaunched
Launched Oct 8 at Gemini at Work 2026: a single general work agent for enterprises — knowledge work, Q&A, content creation, and coding from one prompt box; holds organizational business context, plans tasks, calls skills and tools, connects to business systems, and picks the best model per task (supports Gemini and Claude dual-model routing), with built-in cost controls and enterprise governance; finance/legal editions in preview, government/healthcare/retail to follow
Official Google blog + Reuters; early customers include Shopify and PayPal; Kurian said nearly 500 Cloud customers process over 1 trillion tokens a year
Gemini Skills (formerly Gems)Launched
Reusable custom-instruction system: slash / invocation (moving to @), stackable skills, model auto-matching; Gems has begun its retirement countdown
Gems creation/editing stopped 10-13; personal accounts auto-migrate from 11-17; Workspace 2027-03, education 2027-06; Labs' Opal shuts down 11-17; Gems moves into the Gemini app settings page the same day, fully replaced by Skills from Mar 1, 2027 (enterprise)
Gemini Tasks (formerly Gemini Spark / Task Mode)Launched
Free from Nov 17 for users in the US, Canada, Australia, India, Japan and South Korea: built on Gemini 4 Hydrogen, runs multi-step tasks in the background; US users first get call-making (calling businesses for quotes and bookings), Google Pay checkout, and password-manager login; supports Gmail, Calendar, Drive and 60+ third-party apps; 66 new regional voices and a "keep history but don't train on it" privacy option added; from Nov 18, tasks can also be assigned by text message or email (US first, English only; via Google Messages, iMessage or email; texts are not end-to-end encrypted)
Formerly Gemini Spark / Task Mode (I/O 2026 announcement, then beta for $200 Ultra users only); TechCrunch and others, Nov 17

5 · Subscription monetization layer

Just restructured at I/O 2026
Free → Plus → Pro → UltraLaunched
Plus $4.99 (repriced 2026-06-08 from $7.99) / Pro $19.99 main tier; Ultra launched at $249.99, later split into $100 and $200 tiers
Quotas moved from fixed message counts to compute-based dynamic limits
Bundling strategyLaunched
AI + cloud storage + YouTube Premium bundled; 2TB storage on Pro, 30TB on Ultra
"You pay for storage, AI is free" — removing the psychological barrier to paying
AI Plans model access tighteningLaunched
Effective Oct 9: free users lose Flash / Pro models, keeping only Flash-Lite; AI Plus loses the Pro model from Nov 13 (keeping Flash / Flash-Lite); AI Pro / Ultra keep all models, and add Deep Think (previously Ultra-only)
Official support docs updated; confirmed by Neowin, XDA, and other sources

6 · Hardware layer

On-device + XR, two tracks
Android XR smart glassesIn Preview
Built with Samsung; teased at I/O 2026, details at Samsung Unpacked on July 22; official launch window is fall 2026 (not yet on sale); first gen is screenless audio glasses, display version to come separately
An entry into XR wearables — the ideal vehicle for voice agents
Pixel AILaunched
On-device AI features; the testing ground on Google's own hardware
Paired with the cloud Gemini 3 series

7 · Compute & partnerships

Trading credit for a compute footprint
TeraWulf compute lease credit backingLaunched
Three credit support agreements signed Nov 4, backing Anthropic's 20-year data-center lease in Hawesville, Kentucky (401MW, total lease value ~$19B); if Anthropic defaults or goes bankrupt, Google can take over the lease or pay a termination fee in exchange for equity in the project subsidiary Raylan; backing takes effect in stages as data halls are delivered (end of 2027 to Q2 2028), and ends three months after Anthropic holds an investment-grade rating
The lease itself was signed in July; Google lends no money and builds no campus; rent counts from hall delivery; CoinDesk, Reuters, and The Inference, 11-05 (citing TeraWulf's 8-K filed 11-04)

TimelineMilestones

Radar zone · Unconfirmed (not in the main layers)

Argon scores 53 on the Artificial Analysis AI Index, Gemini 3.8 Flash 41(felloai roundup, single source — no cross-reference from official or mainstream outlets) · Bloomberg says Google insiders doubt Argon's real coding ability and benchmark scores(single source via explainx.ai; Google has rebutted the claim) · Google says Gemini 4 Argon agents freed 300+ TiB of memory in data centers(official Google claim, via a single-source AI news video roundup — no written transcript cross-reference) · Gemini 4 Argon coding benchmark details(Winbuzzer 10-14: DeepSWE v1.1 77.9% / FrontierSWE v2 55.0% / Terminal-Bench 4.0 57.4%, unified 1M context; single source, not cross-verified) · Google Messages pushes Gemini scam-detection Enhanced Protection to Beta users(single-source relay, citing Digital Trends and 9to5Google — no Google official announcement that day). · TeraWulf lease clause "tenant must use only Google-approved TPUs throughout"(seen only in 8-K attachment commentary; CoinDesk and Reuters wires did not mention it; pending verification against the 8-K original, not in the main layers) · Antigravity adds Gemini Live voice mode; NotebookLM Flashcards can add images, timers, and mobile reminders(single source). · Gemini 4 Carbon tops the MathArena math benchmark(Single source: takes the top spot alongside Claude Fable 5.2, replacing GPT-6.1 Sol; not cross-verified)

Verification: Official docs and blogs (antigravity.google, etc.) cross-checked against multiple independent outlets and developer notes; MAU, capex and other figures from third-party I/O 2026 recaps; subscription prices vary by region — check the local page. Unverified entries from the previous edition (rumors written up as launches) have been downgraded or moved to the radar zone per this standard. Gemini 4 Argon confirmed by the official Google blog (10-01) and Reuters, among multiple sources; the Gems→Skills migration cross-checked across unite.ai and others; pricing is the official line; the AI Plans model-access tightening confirmed by official support docs and Neowin, XDA, and other sources (10-03); Nano Banana 2.1 per the official Gemini API changelog (Releasebot 10-06 parse); the Gemini enterprise general agent per the official Google blog and Reuters (10-08); EmbeddingGemma 2 per unite.ai and verdict.co.uk (released 10-06); Gemini 4 Argon global GA per VentureBeat and other tech outlets (10-13), pricing the official line; Gemini 4 Argon embodied/robotics reasoning announced per the official DeepMind blog (10-16); Gemini business tools launched per The Hindu BusinessLine (10-21: business landing pages, Nano Banana 2.1 brand-marketing visuals, spreadsheet/invoice financial analysis; Adobe / Canva / Dropbox / Shopify / Wix / Zoho / Xero integrations; rolling out globally); Gemini on Maps for Apple CarPlay + Personal Intelligence Maps extension per 9to5google and techbriefly (10-22); Gemini Notebook became a Connected App (10-23: manage/query notebooks in natural language mid-conversation, on by default globally, except EEA); apps built in Gemini can be published to Google Sites per official Google Workspace Updates (10-23: apps stay interactive, no code copying). ChromeOS 156 per ghacks.net (10-24: Gemini media understanding on Chromebook Plus, Gemini features in Chrome files/quizzes, talk-to-type dictation, and video summaries); Agent Compute Engine per official Google Cloud announcement (10-22: one hardware-isolated microVM per agent, zero-credential sandbox, customer preview in the coming months); Google Messages Gemini scam-detection Beta is a single-source relay (no Google official announcement that day), now in the radar zone; Genie 4 research preview per the DeepMind announcement and multiple tech outlets (10-27: native audio, real-time scene editing, ~2-minute coherence, SynthID watermarks); Gemini 4 Carbon released per Reuters (10-28: cybersecurity focus, secure-code success rate more than 2x Argon, Fairwind-first opening, $2/$10 intro pricing). The K-12 school policy per TechCrunch, MediaNama, and Gizmodo 11-04 reports (citing Google education lead's 11-03 blog and X clarification). The TeraWulf lease backing per CoinDesk, Reuters, and The Inference 11-05 reports (citing TeraWulf's 8-K filed 11-04). Gemini Omni 1.2 Flash confirmed by the official Google blog, 11-05. The DeepMind RSA-1024 research per the DeepMind research VP's X post and multiple 11-06 cross-reports — the 1,600x figure is a theoretical estimate, the measured result is 3.5x CPU saved on benchmark-key reproduction; Google taking over Crusoe's Wyoming data-center project confirmed by Crusoe on 11-06, price undisclosed. Gemini Memory summary per 9to5Google's 11-09 report; Google AI Plus losing Gemini Pro per Root-Nation's 11-09 report (the change was notified by email in mid-October). Gemini 4 Carbon's broad opening and pricing per Unite.AI and the Google blog, 11-10 (intro $2/$10, $4/$20 after Nov 30; 59 on the Artificial Analysis Intelligence Index v5.0); the MKA benchmark and Antigravity /swarm preview per the Google blog, 11-10; the November Pixel Drop per the official Google blog, 11-10. Gemini Decision 1 and the Decisions API, AlphaGenome 2, and Gemini Robotics 2.5 confirmed by the official Google and DeepMind blogs on 11-11; Gemini 4 Carbon's 60.4% on CursorBench per Cursor, and the AI Mode country list per the Google AI Mode help page updated Nov 11. Gemini 4 Hydrogen's release confirmed by the official Google blog, 11-12 (API pricing $1.50/$7.50, 1M context, 51 on the Artificial Analysis Intelligence Index high tier); Google AI Plus losing Gemini Pro on Nov 13 as planned per the mid-October email disclosed by 9to5Google and NotebookCheck, with the Nov 18 loss of Claude in Antigravity per the same reports. Gemini's Google Wallet connection per Neowin, Nov 14; Gemini 4 Hydrogen's follow-up benchmarks (57% on Terminal-Bench 4.1, 2.5× the task speed of 3.8 Flash) are Google's own claims; GitHub Copilot's retirements and Hydrogen's arrival per GitHub's Nov 13 announcement. Gemini 4 Carbon topping the MathArena math benchmark is single-source and now in the radar zone. Gemini Tasks free rollout cross-reported by TechCrunch and others, Nov 17. Gemini 4 Neon confirmed by the official Google blog, 11-18 (positioning, $3/$15 pricing, 61 on the Artificial Analysis Intelligence Index xhigh tier, replacing Gemini 3.1 Pro and Argon); the Gemini OneGov extension per Nextgov, 11-18 (extended to Feb 28, 2027, $0.47 per agency, second extension); Gemini Tasks text/email channels per tbreak, 11-18 (US first, English only; texts not end-to-end encrypted). Genie 4's arrival on the Gemini API and Google AI Studio confirmed by the official Google blog, Nov 19 (Interactions API, model ID models/genie-4, ~$0.10/second); the Gemini 4 Neon tiered benchmarks per Artificial Analysis, Nov 19, with the model-card framing Google's own. Last updated 2026-11-20. · Privacy · Terms · Contact
← Back to Collection