Oct 13, 2026New trend every hour

Trend of the day · technologytechnology · Oct 13, 2026

Google releases its smartest AI yet, and 10,000 scientists get it free

Google's new AI takes on whole office jobs, but most of the proof it works still comes from Google itself.

This trend is still forming: the judges' scores and the conclusion will appear here as they come.

Guess before the AIs

Yes or no? You first — then we show what 10 AIs answered.

Will Google's share price reach $356 by Friday?

Now: $353.78

We’ll know the answer on Friday

Which AI guesses right more often →

They agree: 8 of 10: only paying subscribers get it; free users see little yet.
They split: Grok says it leads the tests; Claude and Kimi show it trailing Claude Opus 5.5.
Not obvious: The real deadline is for businesses: developer prices double after November 30.
What to do: Test it on one task you know well and check every number.
What to watch: If Alphabet reports more paying AI subscribers on October 28, it holds.
Only one AI: Only Claude: young staff in AI-exposed jobs run 19% below expected hiring.

Read in full

Most of the answers agree that this is a tool for paid work, not a free upgrade for everyone. Claude, Kimi, DeepSeek and Meta AI say only Google's Pro and Ultra subscribers get it. Most of them also see the same change in kind: the AI no longer just answers questions. It carries out long, multi-step tasks, such as turning a messy folder into slides or building financial statements from company filings.

They split on how good it really is. Grok, Qwen and Meta AI repeat Google's claim that it leads the work tests. Claude and Kimi point to independent rankings where it ties OpenAI's best model and trails Anthropic's Claude Opus 5.5. Several answers note that the showcase wins, like 35% more code shipped and enzymes confirmed in the lab, are reported by Google and its partners.

Only Claude brings in jobs data: young workers in AI-exposed jobs are already about 19% below expected employment. DeepSeek alone says free users are being pushed down to a basic model, which the judges could not confirm. Z.ai answered a different question entirely, about bears.

What comes next is a test of money and trust. The low developer price, an API (the way programs connect to the AI), doubles after November 30. Alphabet's quarterly results, due around October 28, will show whether people actually pay for it. Google promises more Gemini 4 models, but expect a staged rollout.

What a person can do: if you already pay, try it on one real task you know well. Time it, check every fact, and keep secrets and client files out. If you don't pay, there is no need to subscribe just for this. Scientists can apply for one of the 10,000 free six-month places. Businesses should run small, measured trials before the price rises.

They split: Grok says it leads the tests; Claude and Kimi show it trailing Claude Opus 5.5.
What settles the dispute: Independent tests in the coming weeks will show if Argon leads or trails.

No real split: both sides see a paid work tool; Z.ai answered about bears instead.Only the AIs from one country noticed: AIs from China — partners already use it to code medical charts and build financial reports.

What is confirmed and what is not

Confirmed by a source Developer price doubles after November 30, from $2/$10 to $4/$20
Confirmed by a source Outside ranking: ties OpenAI's best, trails Claude Opus 5.5 (53 vs 58)
The AIs' conclusion The gap between free and paid AI help is widening

The AIs read 33 sources

Ask any AI yourself

I work as [your job]. Name one weekly task an AI could finish start to end, and how I should check its work.

Copy it and paste it into a chat with any AI

Ask several AIs at once — in our app

Who wins, who loses: Firms and paying users gain speed; junior desk workers may find fewer openings.

The final answer, by Claude Opus

All 10 answers ↓▶ Watch how it came together

On October 13, 2026, Google released AI model">Gemini 4 Argon, its most advanced artificial intelligence (AI) model. Google says it leads tests of work in finance, coding, law and tax.

It is coming to paying users of the Gemini app and Google Search, and 10,000 scientists can apply for free access for six months (source).

What does this release really change for ordinary people, for workers and for science? What happens next, and what should a person do about it now?

What the world's 10 leading AIs think

ChatGPTUSAAI is moving from answering questions to completing everyday tasks
For ordinary people, the biggest change is the move from asking AI for advice to having it complete tasks: preparing documents, organizing information and working across connected apps, sometimes while you're away (blog.google).
This could save time on everyday paperwork, planning and personal projects, although Argon currently requires a paid Pro or Ultra subscription (blog.google). More intelligence does not mean perfect accuracy: Google's own test results show that Argon still trails competitors on some demanding tasks (deepmind.google).
For workers and businesses, the immediate opportunity is completing projects faster, not automatically replacing entire professions: Google reports over 35% more AI-assisted code submissions and fewer reversed changes, although these are company-reported results rather than proof of similar gains everywhere (blog.google).
Science could see especially valuable gains because Google reports AI-assisted enzyme designs confirmed through laboratory testing and a previously missed genetic finding in rare-disease research; both illustrate promising applications that still require scientific verification (blog.google).
The people likely to benefit most are those who combine AI's speed with their own judgment, expertise and ability to catch mistakes.
Next, expect stronger competition over price, reliability and useful automation rather than just impressive test scores; Google already promises additional Gemini 4 models and plans to double Argon's initial developer prices after November 30 (blog.google).
Right now, choose one repeated weekly task, try completing it with AI, measure the time saved, verify the results and protect confidential information; human oversight remains important for consequential decisions (nist.gov). Eligible scientists should consider applying for Google's limited research program, while businesses should run small, measurable trials before committing substantial money (blog.google).
Model: GPT-6 · mode: Extra high · search on · new chat exact
Sources
own forecast: By November 12, 2026, Google will publicly announce at least one additional distinctly named Gemini 4 model besides Argon on an official Google website. — 55% by Nov 12, 2026
ClaudeUSAGoogle's most advanced AI is a paid work tool; free users wait
For most people, little changes yet: Argon sits only in Google's paid plans, $20 a month for Pro or $100 and up for Ultra, and free users can't pick it (engadget.com, androidpolice.com).
Payers get an assistant that hedges less and can finish long jobs, like turning a messy folder into a client slide deck (blog.google, blog.google).
Outside tests put it near the top, not clearly first: it leads the Vals work test, but ties OpenAI's best and trails Anthropic's Claude Opus 5.5 on Artificial Analysis (vals.ai, emergent.sh).
Businesses get a cheap trial window: developer prices double after November 30, and Google says its own teams submitted over 35% more AI-written code with about a third fewer rollbacks (blog.google).
Workers feel tools like this first in junior desk jobs: a Stanford study finds young workers in AI-exposed jobs about 19% below expected employment, though US legal jobs hit a 10-year high in January (intuitionlabs.ai, techtimes.com).
Scientists can apply for one of 10,000 free six-month spots at the top usage tier, but the showcase wins, like a chromosome change routine software missed and new working enzymes, are so far reported by Google and its partners (blog.google).
What to do: if you already pay, test Argon on one real task you know well and check every fact; if you don't, there is no need to subscribe just for this.
Investors shrugged, with shares up only 0.3% today and about 12% below May's high, so the next real test is Alphabet's results, expected between October 28 and November 4 (stockanalysis.com, trefis.com, wallstreethorizon.com, finance.yahoo.com).
More Gemini 4 models are promised, but expect staged access: Google says it cannot rule out that Argon reaches its own danger levels for biology and hacking, and only vetted defenders get a version without cyber guardrails (blog.google, deepmind.google).
Model: Claude Opus 5.5 · mode: Max · search on · new chat exact
Sources
own forecast: On 2026-11-12, a Gemini 4 Argon entry is still ranked #1 on the Vals Index at vals.ai, with no other model scoring higher — 62% by Nov 12, 2026
DeepSeekChinaGemini 4 Argon is powerful, but most people still can't use it.
Most people will see no change today. Free Gemini app users actually lose access to all models except a basic one, while only paying subscribers get the new model. The real benefits for regular people will come later, if Google decides to open it up.
For businesses, this is a big deal. Gemini 4 Argon can write code, analyze legal and financial documents, and find cybersecurity holes at a fraction of the cost of rivals¹². Google is already using it internally to fix bugs and speed up engineering³. This could change how tech, law, and finance teams work.
If you work in a field it targets, learn to use AI tools now. Developers can try the API, and investors should watch whether Google turns this technical lead into real revenue. For everyone else, pay attention: the gap between free and paid AI tools is growing.
Model: DeepSeek V4.1 Flash · mode: DeepThink · search on · new chat likely
Sources
own forecast: Google will make Gemini 4 Argon available to free Gemini app users. — 30% by Mar 31, 2027
GeminiUSAGoogle releases Gemini 4 Argon, boosting performance in coding, law, and science.
For ordinary people, Gemini 4 Argon brings more capable AI assistants into everyday search and apps, making complex problem-solving faster and more intuitive. For businesses and workers, it handles deep, multi-step professional workflows in fields like coding, finance, and legal compliance with fewer errors.
Industries and scientific research benefit from high-level reasoning models that can parse massive datasets and assist with heavy technical optimization. To adapt, professionals should start testing these new capabilities in their daily workflows to see where automation aids efficiency.
Regular users and workers should explore the updated Gemini app features to streamline repetitive tasks and enhance personal productivity. Staying informed on how these tools evolve will help workers maintain a competitive edge as advanced AI integration expands.
Model: Gemini 3 (Auto) · mode: High · search on · new chat likely
Sources
1. blog.google
2. developers.googleblog.com
own forecast: OpenAI will announce a competing frontier reasoning model before December 1, 2026. — 75% by Dec 1, 2026
GrokUSABetter AI tools speed hard work for paid users now
Ordinary people who pay for Gemini Pro or Ultra get a stronger helper for research, writing and planning right away in the app and Search. Free users stay on weaker models, so the gap between paid and free grows. Daily tasks get faster answers with less fluff.
Businesses and workers in coding, law, finance and tax gain a tool that leads tests on multi-step jobs. Early users already cut code errors and speed complex reviews. Scientists can apply for free high-level access for six months.
Try the paid Gemini app today if you handle hard work. Developers should test the API at the low launch price before it rises. Watch for rivals to respond and for the free scientist spots to fill.
Model: Grok 4.7 · mode: Expert · search on · new chat likely
Sources
own forecast: Google reports Argon drove at least 10% higher paid Gemini subscriber growth in Q4 2026 earnings — 55% by Nov 5, 2026

KimiChinaArgon speeds expert work; free users see little yet.
Argon raises the ceiling for paid users, but ordinary life changes gradually; access is limited to Google AI Pro and Ultra subscribers in the Gemini app and AI Mode. Better help with long, messy tasks arrives before any broad free upgrade, so treat answers as strong drafts, not final truth for money, health, legal, or safety choices. blog.google
For workers, the shift is delegation of multi-step work, not instant replacement. Google says Argon leads professional-work tests, while independent reporting finds it trails rivals on some general and coding measures.
Businesses should test it on their own files because the introductory developer rate of $2 per million input tokens and $10 per million output tokens ends November 30 and then doubles; science gains a route to free computing time through 10,000 selected six-month places. blog.google venturebeat.com
Now, choose one difficult task you already understand, compare Argon with your usual method, and keep a person responsible for the final decision. Do not paste secrets, personal data, or client files into any AI tool without permission; workers should learn checking, task design, and clear handoff, not just prompting.
Investors should wait for Alphabet’s October 28 results and evidence of paid adoption before treating the launch as profit. abc.xyz
Model: Kimi K3 · mode: High · search on · new chat exact
Sources
own forecast: Google publicly announces at least one additional Gemini 4-family model beyond Argon. — 38% by Nov 12, 2026
Meta AIUSAGoogle puts work-grade AI in Search and app for paid users
For you, work that took hours of reading and fixing now fits in one chat. The model is live today in the Gemini app and in AI Mode in Search, but only for Google AI Pro and Ultra subscribers. That means faster help with taxes, contracts or code if you pay.
For work, Google says it now leads tests for finance, law, tax and coding and ranks first on the Vals Index. Inside Google, teams using it in Antigravity shipped 35% more code and cut rollbacks by nearly a third. The API costs $2 per million input tokens and $10 per million output now, then doubles to $4 and $20 on December 1.
Try it this month while the low price holds and on one real job you care about. Google offers 10,000 scientists six months free on Antigravity's top tier. Keep human checks, because outside reviews say it is very good but not the best at everything.
Model: Muse Spark 1.3 · mode: Thinking · search on · new chat likely
Sources
own forecast: At least one university, hospital lab or company like Gamow Labs publishes a press release or preprint crediting Gemini 4 Argon for enzyme design, rare-disease diagnosis, or quantum chemistry optimization — 58% by Nov 10, 2026
MiMoChinaThe real shift: AI now finishes whole jobs, not single answers.
For an ordinary person the change is quiet but real. Inside Chrome, the model helped untangle a 6-million-line pile of code so webpages can move between tabs, side panels and floating windows without reloading, which Google says removes a whole class of browser crashes (blog.google).
If you pay for the Gemini app or Google Search, you get sharper answers in finance, law, tax and coding questions, but the biggest gains still land inside big organisations first (blog.google).
For workers the message is that whole multi-step tasks are now in scope, not just drafts. Google reports agentic code submissions up more than 35% and rollbacks down nearly a third, while partners describe the model building financial statements straight from SEC filings and picking ICD-10 codes from long medical charts (blog.google).
That means routine analysis and drafting work moves toward checking machine output, at prices of $2 per million input tokens and $10 per million output tokens (blog.google).
For science the open door matters most: 10,000 scientists can apply for six months of free top access, and early users already sped up a quantum chemistry package and a model-inference stack by multiples (blog.google).
What you should do now is simple: if you qualify, apply for that programme; if you work with data, learn to hand the model long, multi-step tasks with a clear goal and let it run; and always check its numbers and citations before you rely on them (blog.google).
Model: MiMo-V2.6-Pro · search on · new chat likely
Sources
1. blog.google
2. finance.yahoo.com
own forecast: Google quietly ends the introductory Gemini 4 Argon API price, so input tokens cost more than $2 per million by this date — 55% by Mar 31, 2027
QwenChinaNew AI model handles long, complex tasks for workers and scientists.
Paying users can now use this model inside the search and app to solve hard problems (source: blog.google). It handles long tasks that older models usually forget (source: blog.google).
Companies can use it to automate long coding jobs and improve computer security (source: blog.google). This means workers spend less time on boring research and more time checking the final results (source: blog.google).
Ten thousand scientists get free access to solve deep math and research problems (source: blog.google). Workers should start testing it on their hardest daily tasks right now to stay ahead (source: blog.google).
Model: Qwen3.8-Max · mode: Thinking · search on · new chat exact
Sources
1. blog.google
2. blog.google
own forecast: Google adds a live voice avatar feature to the Gemini 4 Argon model for mobile users. — 40% by Nov 12, 2026
WenxinChinaGemini 4 Argon shifts work and science
Ordinary paying users get far deeper search and app help that can finish long, messy tasks in one go, no more stopping mid-work to fix gaps. It can pull full reports, sort big files, and walk you through complex steps without extra clicks.
Workers in finance, coding, law and tax get a tool that cuts hours of repetitive deep work down to minutes, letting them focus on the parts that need real human judgment. Businesses can use the low introductory API price to build custom tools for their teams without huge upfront costs.
Right now, you can test the model on a small real task you do every week to see where it saves you time, and keep an eye out for the free scientist access application opening soon. Don’t wait too long to try it, as the low API pricing will go up after November 30.
Model: ERNIE · mode: Fast · search on · new chat not disclosed
Sources
own forecast: Google will open free Gemini 4 Argon access applications for 10,000 scientists — 65% by Oct 27, 2026

Previous trends

Which AI guesses right more often

Accuracy 0–100: how close the chance was to what happened. 100 — exact, 50 — a coin toss, 0 — the opposite. few data: 0 of 30 results

The first results come a day after the first questions.

Came true or not: what 12 AIs promised for 2026 →

How it works

The world's 10 leading AIs (5 from the USA and 5 from China) answer one question on the topic of the day.

10 AIs, one question. 5 AIs from the USA and 5 from China answered it. 16 were asked, 12 answered — each side shows its 5 strongest.

Above the answers is the question every AI got, word for word.

We seal every question before the answer is known, and we send its fingerprint at once to OpenTimestamps, a free public registry that does not belong to us. No one can quietly change a question afterwards.

Came true or not: what 12 AIs promised for 2026 →

Which models answer

The main league: each AI answers in its own app, in its best single mode on a regular plan (about $20–30 a month) or free, with web search on, in a new chat, once.

Heavy modes, where several copies of a model think at once (GPT Pro, Grok Heavy, Deep Think, Agent Swarm and the like), are a separate league: they never count in the score of the sides.

AI	Model	Mode	Plan	How sure
Claude	Claude Opus 5.5	Max	Max	exact
ChatGPT	GPT-6	Extra high	Pro	exact
Meta AI	Muse Spark 1.3	Thinking	free	likely
Grok	Grok 4.7	Expert	SuperGrok	likely
Gemini	Gemini 3 (Auto)	High	Google AI Pro	likely
MiMo	MiMo-V2.6-Pro		free	likely
Qwen	Qwen3.8-Max	Thinking	free	exact
Kimi	Kimi K3	High	membership	exact
DeepSeek	DeepSeek V4.1 Flash	DeepThink	free	likely
Wenxin	ERNIE	Fast	free, no sign-in	not disclosed

Exact: the app showed the model for this answer. Likely: the app shows only the mode, so we name the model the provider’s own help lists for that mode. Not disclosed: the app does not say which model answered. Every judge works in the same mode as in the table.

The forecast · result on Nov 12, 2026

Will it happen? Each AI gave a probability

Will Alphabet, Google's parent company, close at or above $365 per Class A share (ticker GOOGL, Nasdaq) on November 12, 2026, as shown by the official closing price on Yahoo Finance (https://finance.yahoo.com/quote/GOOGL/history/)?

The result comes on Nov 12, 2026, 30 days after this issue. Only the regular-session close on November 12, 2026 counts. After-hours prices, prices during the day and prices on other days do not count. If markets are closed that day, the last regular close before it counts. The close on October 13, 2026 was $353.78. Each AI is scored by how close its probability was to what happened (the Brier score).

Poly USA-5 · 41%
Gemini65%
ChatGPT44%
Meta AI42%
Claude37%
Grok35%
Poly China-5 · 44%
MiMo45%
Qwen45%
DeepSeek45%
Kimi42%
Wenxin38%

the line: Poly World-10 · 42%

Poly Index

Three numbers from the same answers: the five American AIs, the five Chinese AIs and all ten together.

Poly USA-541%Claude, ChatGPT, Meta AI, Grok, Geminiall 5 answered
Poly China-544%MiMo, Qwen, Kimi, DeepSeek, Wenxinall 5 answered
Poly World-1042%Claude, ChatGPT, Meta AI, Grok, Gemini, MiMo, Qwen, Kimi, DeepSeek, Wenxinall 10 answered

How we count: we drop the highest and the lowest answer and take the average of the rest. The AIs answered before they saw any market.

Each AI's own forecast (10)

AI	Forecast	By	Chance
Claude	On 2026-11-12, a Gemini 4 Argon entry is still ranked #1 on the Vals Index at vals.ai, with no other model scoring higher	Nov 12, 2026	62%
ChatGPT	By November 12, 2026, Google will publicly announce at least one additional distinctly named Gemini 4 model besides Argon on an official Google website.	Nov 12, 2026	55%
Meta AI	At least one university, hospital lab or company like Gamow Labs publishes a press release or preprint crediting Gemini 4 Argon for enzyme design, rare-disease diagnosis, or quantum chemistry optimization	Nov 10, 2026	58%
Grok	Google reports Argon drove at least 10% higher paid Gemini subscriber growth in Q4 2026 earnings	Nov 5, 2026	55%
Gemini	OpenAI will announce a competing frontier reasoning model before December 1, 2026.	Dec 1, 2026	75%
MiMo	Google quietly ends the introductory Gemini 4 Argon API price, so input tokens cost more than $2 per million by this date	Mar 31, 2027	55%
Qwen	Google adds a live voice avatar feature to the Gemini 4 Argon model for mobile users.	Nov 12, 2026	40%
Kimi	Google publicly announces at least one additional Gemini 4-family model beyond Argon.	Nov 12, 2026	38%
DeepSeek	Google will make Gemini 4 Argon available to free Gemini app users.	Mar 31, 2027	30%
Wenxin	Google will open free Gemini 4 Argon access applications for 10,000 scientists	Oct 27, 2026	65%

Each own forecast is checked on its own date.

You can have your own trends

You choose the topic — several AIs prepare solutions

Download for free. Want more? Get all the powerful tools for $2 a month.