Timeline, August–October 2026
Dates come from Google’s own blog, the Gemini app release notes, the Gemini API changelog and the Antigravity blog, re-checked October 28. Benchmark numbers are Google’s own and are quoted only to show what Google chose to emphasize.
| Date | Launch | Why it matters |
|---|---|---|
| August 11 | Gemini app passes 1 billion monthly users | Google calls it its fastest-growing product ever; 100M+ active users on iOS |
| August 12–August 21 | Antigravity: custom agents, IDE extensions, Remote Control | Drive agent sessions on any of your machines from a browser |
| August 13 | Gemini 3.7 Flash GA | Time-limited launch price, aimed at coding and agents |
| August 19 | Student hub; a year of AI Pro for US students, AI Plus in 140+ countries | Free years now, paying users later |
| August 27 | Gemini Omni Flash (video) GA | Video generation in the API; Flow credits in plans |
| September 2 | Gemini 3.8 Flash and 3.8 Flash Cyber | Third Flash in six weeks; Cyber only for trusted defenders |
| September | 3.8 Live, 3.8 TTS, Lyria 3.5, Googlebook pre-orders | Voice and music models refreshed together |
| September 10 | Gemini app for Windows (Alt + Space) | Desktop coverage after macOS |
| September 30 | Skills in Gemini chat, replacing Gems | Reusable instruction packs, as in Claude and ChatGPT |
| September 30 | Gemini 4 Argon announced, Fairwind Program first | Output limit raised from 64K to 1M tokens |
| October 6 | Nano Banana 2.1 GA | Image model refresh, also used for small-business brand assets |
| October 13 | Argon broadly available; Antigravity inside the Gemini macOS app; chat-to-tasks; 40+ connected apps | Google’s biggest joint model and product release of the year |
| October 15 | Universal Cart (US) | One cross-retailer cart for Search and the Gemini app |
| October 20 | Gemini small-business tools; Android Auto worldwide | Distribution to merchants and cars |
| October 21 | OpenAI announces its models and Codex for Google Cloud | Later also in Antigravity and Gemini Agent |
| October 26–October 27 | API stores new projects’ requests by default; gemini-api CLI beta; Genie 4 research preview | Developers should check data-retention settings |
Gemini 4 Argon: a staged release, and a launch price that ends November 30
Argon is the core of Google’s quarter. On September 30 it went only to trusted cyber defenders in the Fairwind Program, and Google said it was taking part in the US government’s voluntary pre-release access process. On October 13 it opened to the API, AI Pro and Ultra, and Fairwind expanded to hundreds more partners. Trusted defenders get a version without cyber guardrails.
- Positioning: long, multi-step work in software engineering, knowledge work such as finance and law, and cyber defense. Google reports DeepSWE v1.1 77.9%, Zapier’s AutomationBench 51.3% and the top Vals Index rank.
- Output limit: 1M tokens (was 64K), with a new continuation protocol in the API for very long outputs.
- Internal use: the Antigravity blog says over 100K Alphabet users a month; over four weeks, 35% more agentic code and 30% fewer post-submit rollbacks.
- API price: $2 / $10 per million tokens (cached input 95% off) only through November 30; $4 / $20 from December 1. There is no free tier for Argon.
- Safety: Google says Argon leads Gray Swan’s indirect prompt injection benchmark, and it runs a misalignment monitor over reasoning and actions that pauses for human confirmation before irreversible steps.
For buyers, the most direct change is that the $19.99 AI Pro now includes the flagship, with lower limits than Ultra. For developers, the key date is December, when the price doubles; for similar jobs 3.8 Flash at $0.75 / $3.75 remains Google’s low-cost option.
The Gemini app and Google AI plans: same prices, different contents
We re-checked the Gemini subscriptions page and Google’s US price feed on October 28; prices match October 19. What changed is what each plan includes:
| Plan | Monthly (US) | Gemini app | Coding | Other |
|---|---|---|---|---|
| Free | $0 | Flash-Lite model | Antigravity free tier (basic weekly limits) | 15 GB storage |
| AI Plus | $4.99 | Flash model, 2x Free limits | — | 400 GB storage; 160+ countries and territories |
| AI Pro | $19.99 | Argon plus Pro / Flash models, 4x Free limits | Entry Antigravity limits (with Argon), higher Jules limits | 5 TB storage, YouTube Premium Lite |
| AI Ultra 5x | $99.99 | 5x AI Pro; Tasks first (select countries) | Higher Antigravity limits, highest Jules limits | 20 TB storage, YouTube Premium, Project Genie |
| AI Ultra 20x | $199.99 | 20x AI Pro | Highest Antigravity limits | 30 TB storage |
In the app, Gemini filled out the desktop (Windows on September 10, Antigravity inside the macOS app on October 13), swapped Gems for skills (September 30), added a Customize menu and 40+ connected apps (mostly US English), and put chat and tasks on one toggle: choose tasks and Gemini keeps working across Google apps while you are away. Limits still refresh every 5 hours up to a weekly cap (Google Help, read by us October 19).
Antigravity and coding tools: Google bets on one line
If you last looked at Google’s coding tools when they meant Gemini CLI and Jules, this is the biggest change. Google said in May it would focus on Antigravity; since June 18, free users and AI Pro / Ultra can no longer use Gemini CLI and must move to Antigravity CLI, written in Go, which keeps Skills, Hooks, Subagents and extensions (now called plugins). Gemini CLI still ships (v0.66.0, October 27) but only for enterprise Code Assist licences, and its README says it is not optimized for Gemini 4.
- The Antigravity line: desktop app, CLI, IDE extensions (August 20), SDK, and the entry inside the Gemini macOS app.
- Remote Control (August 21): connect to and drive Antigravity sessions running on different machines from a browser, with push notifications when an agent needs input. Same idea as remote and cloud sessions in Claude Code and Codex.
- Custom agents: custom agents (August 12), packageable in Google plugins since September 28.
- Billing: a free individual tier with basic weekly limits; AI Pro adds limits and a flexible AI credit pool, Ultra more; businesses via Google Cloud with Gemini Enterprise Standard / Plus from $30 per seat per month, or pay-as-you-go with no seat fee.
- Jules: still on the plans page (higher limits on AI Pro, highest on Ultra), but its public changelog stops at March 9.
Compared with Codex and Claude Code, Google sells coding as a plan benefit priced by AI Pro / Ultra multiples, the same logic as ChatGPT Pro and Claude Max. For a three-way comparison see our batch article AI coding agents: Q4 2026 trends.
API and developers: launch prices, auto-routing, storage by default
- Cadence: 3.6 Flash (July 21), 3.7 Flash (August 13) and 3.8 Flash (September 2) in six weeks; since October 8, 3.7 Flash and 3.5 Flash requests route automatically to newer models.
- Prices: 3.8 Flash $0.75 / $3.75 (through December 31, then $1.5 / $7.5); Argon above. Search and Maps grounding: 5,000 free requests a month, then $14 per 1,000.
- New tools: the Antigravity Agent in the API (new version September 17 with changed tool parameters), managed agents (environment compute not billed in preview), and the gemini-api command line beta (October 27).
- Data retention: since October 26, new paid-tier projects store Generate Content requests by default; existing projects keep their setting. Teams with sensitive data should set store=false or turn logging off.
- Media: Omni Flash video (August 27), Lyria 3.5 music, 3.8 Live voice, 3.8 TTS and Nano Banana 2.1 images all reached general availability this quarter.
Distribution and partners: Google’s biggest card
The Gemini app passed 1 billion monthly users on August 11 (Google has not published how many pay). In three months Gemini reached the Windows desktop, Android Auto and cars with Google built-in (October 20; with permission it can draw on Gmail, Calendar, Photos and more), a US shopping cart (October 15) and small-business tools (October 20; Google says tens of millions of small business owners use Gemini).
Another signal comes from a rival: on October 21 OpenAI announced that its models and Codex are coming to Google Cloud Model Garden and later to Google Antigravity and Gemini Agent, with general availability planned within weeks. A Google Cloud that resells a competitor’s models wants to be the enterprise platform, not only the Gemini vendor.
Six trend calls for the next 6–12 months
These are our calls, not Google’s claims. Each lists the signals we observed, a confidence word (likely / possible / speculative) and what would prove it wrong. Nothing here is investment advice. For the industry-wide view see AI industry trends Q4 2026.
1. Frontier API prices move to “launch price, then list price”, with Flash holding the low end
Our read: Google no longer parks its flagship at a low price for long: it launches cheap to win developers, then returns to list. The real price fight happens in the Flash line. Teams paying per token should budget on post-December prices. Confidence: likely.
- Gemini 4 Argon is $2 / $10 per million tokens only through November 30, doubling to $4 / $20 from December 1 (Gemini API pricing page).
- 3.8 Flash costs $0.75 / $3.75 through December 31, then $1.5 / $7.5; 3.7 Flash also launched on a time-limited price.
- Since October 8, requests to 3.7 Flash and 3.5 Flash are routed to newer models, so old versions do not linger as cheap options.
What would prove it wrong: Google extends Argon’s introductory price past November 30, or sets the list price at $2 / $10.
2. Google’s coding tools converge on Antigravity
Our read: Within a year, an individual developer on Google’s side will code almost only in Antigravity (desktop app, CLI, IDE extensions, SDK, the entry inside the Gemini app). Gemini CLI stays for enterprise Code Assist and Jules fades. Confidence: likely.
- The Gemini CLI README says free users and AI Pro / Ultra moved to Antigravity CLI on June 18, and that Gemini CLI is “not optimized” for Gemini 4.
- Argon reached the Antigravity desktop app and CLI on day one (October 13), the same day Antigravity was built into the Gemini macOS app.
- Jules’s public changelog stops at March 9, while the Antigravity blog shipped every month from August to October (IDE extensions, Remote Control, custom agents in plugins).
What would prove it wrong: Jules gets Argon and resumes regular releases, or Gemini CLI reopens to AI Pro / Ultra.
3. AI Pro vs Ultra becomes “how much”, not “whether”
Our read: The newest model reaches the $19.99 AI Pro quickly; Ultra sells multiples of usage, first access and bigger coding quotas. That is the same ladder logic as ChatGPT and Claude. Confidence: likely.
- Argon arrived for AI Pro and Ultra together on October 13 (Gemini app, AI Mode); the subscriptions page says “Access” for Pro and “Highest access” for Ultra.
- In Antigravity, AI Pro has “entry” limits and Ultra 5x / 20x “higher” / “highest”; both get Argon.
- Tasks and Genie 4 are both listed as Ultra-first.
What would prove it wrong: Google keeps a flagship model or core feature Ultra-only for the long term, or drops the Ultra 5x rung.
4. Google’s strongest models keep going through trusted-access programs first
Our read: Google’s next frontier model (or Argon’s successor) will again go to Fairwind-style programs first and reach paid API customers and subscribers weeks later. Confidence: likely.
- Argon went to Fairwind cyber defenders on September 30 and broadly on October 13, when Fairwind also expanded to hundreds more partners.
- 3.8 Flash Cyber is Fairwind-only, and only trusted defenders get Argon without cyber guardrails.
- Genie 4 (October 27) started with a limited group of trusted testers.
What would prove it wrong: Google’s next flagship opens to every paid API customer and Ultra subscriber on day one.
5. Distribution is Google’s edge: Gemini shows up in more front doors, and Google Cloud resells rivals
Our read: Google competes on reach as much as on benchmarks: phones, cars, Windows, shopping and cloud. We expect OpenAI models to be selectable in Antigravity or Gemini Agent before the end of Q1 2027. Confidence: possible.
- The Gemini app passed 1 billion monthly users (August 11) with 100M+ on iOS; the Windows app (September 10); Android Auto worldwide (October 20).
- Universal Cart in Search and the Gemini app in the US (October 15); Gemini small-business tools (October 20).
- OpenAI announced its models and Codex for Google Cloud Model Garden, then Antigravity and Gemini Agent (October 21).
What would prove it wrong: OpenAI’s general availability on Google Cloud keeps slipping, or Antigravity still offers only Google models in March 2027.
6. For agents, “store by default” becomes the API norm
Our read: Agents need replayable logs and session state, so more APIs will store requests by default and leave opting out to you. Teams handling sensitive data should watch this setting. Confidence: speculative.
- Since October 26, new paid-tier Gemini API projects store Generate Content requests by default (store=true), matching the Interactions API; you can opt out per request or in settings.
- Google shipped the gemini-api command line for coding agents and CI (October 27), running on the Interactions API and managed agents.
- Environment compute for managed agents and the Antigravity Agent is not billed during the preview, which pulls work into Google’s session model early.
What would prove it wrong: Google reverses the default, or other major labs commit to not storing by default.
Choosing between Gemini, ChatGPT and Claude
ClaudeMax does not sell Google AI plans. If you already live in Google’s ecosystem (Gmail, Drive, YouTube Premium, large storage), the AI Pro or Ultra bundle is hard to replace; subscribe with Google directly. If you mainly code, research or run agents and are deciding between Gemini and ChatGPT or Claude, these are the same-class options we sell (live prices):
| If you… | Buy | Price here (30 days) |
|---|---|---|
| were going to buy AI Ultra 5x and mostly code or run agents in Codex and ChatGPT Work | ChatGPT Pro 200 top-up (ready-made also available) | ¥1,300 ≈ 193.75 USDT; ready-made ¥1,349 ≈ 201.05 USDT |
| were going to buy AI Ultra 20x, are one very heavy user, or need Ultrafast | ChatGPT Pro 500 top-up (ready-made also available) | ¥3,299 ≈ 491.66 USDT; ready-made ¥3,399 ≈ 506.56 USDT |
| want Claude Code for coding and long-form writing | Claude Max 5x or Max 20x top-up | Max 5x ¥999 ≈ 148.89 USDT; Max 20x ¥1,999 ≈ 297.92 USDT |
| want more than Max 5x per session in a separate account | Claude Max 6.25x ready-made: a Claude Team-plan Premium seat, 6.25x Pro per session (Anthropic support article 9266767) | ¥1,379 ≈ 205.52 USDT |
| just want a stronger everyday assistant (the AI Pro slot) | ChatGPT Plus or Claude Pro top-up | Plus ¥185 ≈ 27.58 USDT; Claude Pro ¥249 ≈ 37.11 USDT |
For a line-by-line comparison of the two $200-class plans read ChatGPT Pro 200 vs Google AI Ultra; for OpenAI’s and Anthropic’s quarters, OpenAI in October 2026 and Anthropic Claude in October 2026; for Meta, our batch article Meta Muse in October 2026.
How buying here works: top-ups are added to your own account and your sign-in data (Session JSON for ChatGPT, a cookie export for Claude) is checked before you pay; if the check fails you pay nothing. Ready-made accounts are hand-delivered in about 30 minutes to 4 hours after you verify a delivery email. Pay in USDT on TRC20 or BEP20. Ready-made accounts are not refundable once delivered; if anything goes wrong, contact the bottom-right live chat or Telegram @ccnb1937 promptly.
Risk, briefly: each vendor’s terms apply per user, and access from unsupported regions may be restricted. ClaudeMax is independent and is not Google, OpenAI, Anthropic or an authorized reseller; vendors set the allowances and may change them again.