Back to the trend of the day: ChatGPT's maker is reported to be worth $1.25 trillion after new funding

Poly Rating · from the trend of the day

Most Valuable AI Companies 2026: Price vs Published Revenue

Anthropic's SEC filing shows about $67.8B of annualised revenue behind its $965B price, about 14x. OpenAI's larger $1.25T price rests on about $50B that it told investors and has never filed, about 25x.

No. 1: Anthropic
10 places
Current as of November 9, 2026, 18:00 EST

01Final list

The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.

Order: from most to least AI agreement.

A bright badge means this AI put the entry on its list. A dim badge means it did not.

1
Anthropic
The only lab with revenue in an SEC filing: $16.96B in Q3, about $67.8B a year. That puts its $965B May price at about 14x.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 1 · Nemotron: 2 · Nova: 3 · Claude: 1 · ChatGPT: 1 · StepFun: 2 · DeepSeek: 2 · Kimi: 2
What each AI noted 8
- Metaplace 1
  Filed $33.2B in nine months - the only lab whose revenue is in an S-1 - at about 14x Q3-annualised sales.
  Source ↗
- Nemotronplace 2
  $40B valuation / $1B ARR = 40× multiple; $800M ARR confirmed in company memo, stronger disclosure than peers
- Novaplace 3
  $965B valuation on $47B revenue run-rate (20.5x) after $65B Series H
  Source ↗
- Claudeplace 1
  About 15x its $65B run rate at the $965B May price, and the one lab whose revenue is backed by a (confidential) IPO filing. At the $1.5T IPO target it would be about 23x.
  Source ↗
- ChatGPTplace 1
  A $965 billion financing price now sits beside $67.8 billion of annualized quarterly revenue in an actual SEC filing.
- StepFunplace 2
  Roughly a seventh of OpenAI's price on a revenue number it is far less shy about.
  Source ↗
- DeepSeekplace 2
  $965B last private mark against $67.2B annualised run-rate is 14.4x; the IPO target of $1.5T would make it 22.3x.
- Kimiplace 2
  The only frontier lab with regulator-filed revenue - $17.0B in Q3 alone, up 1,127% y/y - yet IPO buyers are reportedly weighing $1.1T, not the $2T hoped for a month earlier.
  Source ↗
8 AIs put it here
Source ↗
2
OpenAI
It has the freshest and biggest price: $1.25T, $150B below the $1.4T it first sought. That is about 25x a roughly $50B run rate told to investors and never filed.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 3 · Nemotron: 1 · Nova: 4 · Claude: 2 · ChatGPT: 3 · StepFun: 1 · DeepSeek: 1 · Kimi: 1
What each AI noted 8
- Metaplace 3
  $1.25T on $50B told to investors, not filed - the freshest price, the thinnest paper.
  Source ↗
- Nemotronplace 1
  $1.25T valuation / ~$4B ARR = 312× multiple; only ~$1.6B of ARR confirmed in company blog, rest from The Information reporting
- Novaplace 4
  $852B valuation on $25B revenue target (34x) with 900M weekly ChatGPT users
- Claudeplace 2
  A fresh $1.25T price, down from the $1.4T–$1.5T first sought, on about $50B of run rate: roughly 25x, with no public filing behind it.
  Source ↗
- ChatGPTplace 3
  The $1.25 trillion headline is about 25 times comparable reported annualized sales, and $150 billion below the valuation initially sought.
  Source ↗
- StepFunplace 1
  A $1.25 trillion price resting on a revenue number the company has never filed.
  Source ↗
- DeepSeekplace 1
  $1.25T price on ~$50B of self-stated revenue is 25x — and the “markdown” from $1.4T actually raised the price per dollar of revenue.
  Source ↗
- Kimiplace 1
  Wired $11B of a ~$20B round at a $1.25T post-money price - 11% below the $1.4T it first sought - while the ~$50B revenue behind its ~25x multiple is a number investors were told, never filed.
  Source ↗
8 AIs put it here
Source ↗
3
Databricks
Its company-stated run rate of more than $7B, growing over 80%, puts the $190B price at about 27x. The company published the figure, but it is unaudited and not filed.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 2 · Nemotron: 3 · Nova: 2 · Claude: 3 · ChatGPT: 2 · StepFun: 5 · DeepSeek: 4 · Kimi: 4
What each AI noted 8
- Metaplace 2
  Company itself said over $7B run-rate - 27x sales, the cleanest company-stated multiple after Anthropic.
  Source ↗
- Nemotronplace 3
  $55B valuation / $2.4B ARR (filed) = 23× multiple; only AI-first co. with full GAAP filing via confidentially submitted S-1
- Novaplace 2
  $134B valuation on $7B+ annualized revenue (19x) with 80% YoY growth
  Source ↗
- Claudeplace 3
  The cleanest numbers in the list: $190B on a stated $7B run rate, about 27x, growing over 80%.
  Source ↗
- ChatGPTplace 2
  More than $7 billion in annualized revenue supports a $190 billion price, with enterprise customers paying for data infrastructure as well as AI.
  Source ↗
- StepFunplace 5
  The only top-ten name that puts its own revenue number in a press release.
  Source ↗
- DeepSeekplace 4
  $190B at $7B run-rate is 27.1x — the steepest multiple among the profitable-at-scale data/AI platforms.
  Source ↗
- Kimiplace 4
  $190B at 27x a self-reported $7B run rate - priced like a model lab while free-cash-flow positive, on numbers a public competitor's auditors would never sign off.
  Source ↗
8 AIs put it here
Source ↗
4
Scale AI
The $29B is Meta's June 2025 price, now 17 months old. On the roughly $1B revenue the CEO told Forbes, that is 29x, and one secondary-market estimate puts it at $15.7B.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 4 · Nemotron: 5 · Nova: 1 · Claude: 4 · ChatGPT: 6 · StepFun: 8 · DeepSeek: 9 · Kimi: 7
What each AI noted 8
- Metaplace 4
  A $29B price frozen since June 2025, now chasing just over $1B of company-told revenue.
  Source ↗
- Nemotronplace 5
  $14B valuation / $800M ARR (company) = 17.5× multiple; profitable per company statement, rare in cohort
- Novaplace 1
  $29B valuation on $2B projected revenue (14.5x) with Meta’s $14.3B stake
  Source ↗
- Claudeplace 4
  A $29B price set by Meta's 2025 stake, so the price is 16 months old.
  Source ↗
- ChatGPTplace 6
  A $29 billion headline financing value contrasts with a $15.7 billion estimated secondary-market value.
- StepFunplace 8
  The only top-ten name whose majority owner is a strategic buyer, not a fund.
  Source ↗
- DeepSeekplace 9
  $29B from Meta’s 2025 stake, against ~$1B of 2026 revenue guidance, is ~29x — a stale price sitting next to fresh 2026 rounds.
  Source ↗
- Kimiplace 7
  ~$29B is a 17-month-old strategic mark set by Meta buying 49% - the same deal that drove away Google, OpenAI and xAI, the very revenue being priced.
8 AIs put it here
Source ↗
5
xAI (SpaceX AI)
Its 2025 AI revenue of $3.2B is filed inside SpaceX's S-1, but the $250B price ended with the February merger. On that base the price is about 78x, and the segment mixes X ads with compute rental.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 6 · Nemotron: 4 · Nova: 5 · Claude: 7 · ChatGPT: 4 · StepFun: 4 · DeepSeek: 3 · Kimi: 3
What each AI noted 8
- Metaplace 6
  Filed $3.2B in 2025 inside SpaceX's S-1, but its $230B standalone price died in the February merger.
  Source ↗
- Nemotronplace 4
  $50B valuation / $100M ARR = 500× multiple; revenue almost entirely third-party estimates, no company disclosure
- Novaplace 5
  $250B embedded valuation inside SpaceX with $500M standalone revenue (500x)
- Claudeplace 7
  Has no standalone price since SpaceX absorbed it on 2 Feb 2026; sources quote $230B, $50B and about $500M revenue.
  Source ↗
- ChatGPTplace 4
  The $250 billion transaction value belongs to the AI business acquired by SpaceX, not to a standalone company still raising independently.
  Source ↗
- StepFunplace 4
  A price that assumes Colossus turns into a business before the building depreciates.
- DeepSeekplace 3
  $230B standalone mark on $10.24B annualised SpaceX-filing revenue is 22.5x — but on the old ~$500M standalone run-rate it was ~460x.
  Source ↗
- Kimiplace 3
  Marked at $250B when SpaceX absorbed it in Feb 2026 - and now earns more renting Colossus to rivals (~$1.6B of Q2 revenue from Anthropic and Google) than from Grok itself.
  Source ↗
8 AIs put it here
Source ↗
6
Perplexity
Last closed at $20B in September 2025. Its revenue, about $750M, is reported only by others: 27x at the closed price, 40x at the $30B-plus talks.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 5 · Nemotron: 6 · Nova: 7 · Claude: 5 · ChatGPT: 7 · StepFun: 7 · DeepSeek: 8 · Kimi: 10
What each AI noted 8
- Metaplace 5
  $20B price is 14 months old, while reported revenue tripled to $750M - the price is stale, the growth is not.
  Source ↗
- Nemotronplace 6
  $9B valuation / $100M ARR = 90× multiple; $50M ARR confirmed in company tweet, rest from leaks
- Novaplace 7
  ~$20B valuation with answer-engine revenue not publicly broken out
  Source ↗
- Claudeplace 5
  About $21–23B against a revenue target (not a run rate) of $656M for 2026: roughly 32x–35x.
  Source ↗
- ChatGPTplace 7
  An estimated $750 million revenue run rate is growing against a $20 billion completed price, while a higher $30 billion-plus financing was only under discussion.
  Source ↗
- StepFunplace 7
  A higher price-to-revenue than OpenAI on reported numbers, for a product that is still a search box.
  Source ↗
- DeepSeekplace 8
  A reported >$30B price on >$750M annualised revenue is ~40x — the multiple barely moved even as revenue tripled.
  Source ↗
- Kimiplace 10
  $20B last mark with $30B IPO-talk against a ~$750M run rate - roughly 27-40x revenue for an answer engine still renting CoreWeave clusters to keep up.
  Source ↗
8 AIs put it here
Source ↗
7
Mistral AI
Over €21B (about $24B), set in September, on a $400M run rate stated by its CEO. That is about 60x, or 24x if the $1B year-end target is reached.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 8 · Nemotron: 7 · Nova: 6 · Claude: 6 · ChatGPT: 5 · StepFun: 6 · DeepSeek: 10 · Kimi: 9
What each AI noted 8
- Metaplace 8
  Over €21B on a €1B forecast that has not happened yet - the multiple is a promise, not a filing.
  Source ↗
- Nemotronplace 7
  $6B valuation / $50M ARR = 120× multiple; €50M ARR stated in French regulatory filing (AMF), strong disclosure for EU co.
- Novaplace 6
  $14B valuation on $100M ARR (140x) with open-weight model strategy
  Source ↗
- Claudeplace 6
  The highest estimated multiple with any revenue figure: about 60x on a roughly $400M estimate.
  Source ↗
- ChatGPTplace 5
  Europe's €21 billion-plus frontier-model company has a roughly $400 million revenue estimate, making its price unusually sensitive to growth.
  Source ↗
- StepFunplace 6
  Europe's price tag on a revenue number the company only gestures at.
  Source ↗
- DeepSeekplace 10
  $24B at $400M trailing ARR is ~60x; on the company’s $1B 2026 target it is ~24x — the gap is the whole growth bet.
  Source ↗
- Kimiplace 9
  Europe's champion at ~$24B post-money - about 24x the $1B revenue it hopes to reach by year-end, and ~60x the $400M run rate it last actually stated.
8 AIs put it here
Source ↗
8
Thinking Machines Lab
A $40B pre-money round was reported closing in November but is not yet confirmed. On a reported $100M-plus run rate that is about 400x, against 120x at its $12B seed.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 9 · Nemotron: 9 · Nova: 8 · Claude: 9 · ChatGPT: 8 · StepFun: 10 · DeepSeek: 6 · Kimi: 5
What each AI noted 8
- Metaplace 9
  $40B talks on over $100M reported revenue - 400x sales, the thinnest revenue base in the top 10.
  Source ↗
- Nemotronplace 9
  $9B valuation / $10M ARR = 900× multiple; minimal revenue from Mira Murati's stealth launch, valuation on team pedigree
- Novaplace 8
  $12B valuation from $2B seed round with no product revenue yet
  Source ↗
- Claudeplace 9
  Last priced at about $12B on roughly $2B raised in 2025, with no 2026 revenue found.
  Source ↗
- ChatGPTplace 8
  A proposed $40 billion pre-money price is roughly 400 times a reported $100 million-plus revenue run rate, before the round has closed.
  Source ↗
- StepFunplace 10
  $10B for a seed round and a research blog.
  Source ↗
- DeepSeekplace 6
  $40B pre-money on just over $100M annualised revenue is a ~400x multiple — one of the highest ever seen in the AI cycle.
  Source ↗
- Kimiplace 5
  Closing $5B at a $40B pre-money price - over 400x a reported >$100M run rate - nine months after its own $50B round talks collapsed.
  Source ↗
8 AIs put it here
Source ↗
9
Safe Superintelligence (SSI)
Valued at $32B (April 2025) with about 50 staff, no product and no revenue, so there is nothing published to rest on. Nvidia's $5B in July came at undisclosed terms.
Included it: Meta, Nemotron, Nova, Claude, ChatGPT, StepFun, DeepSeek, Kimi. Did not include: Wenxin, Qwen.
In their lists: Meta: 10 · Nemotron: 8 · Nova: 9 · Claude: 10 · ChatGPT: 9 · StepFun: 9 · DeepSeek: 7 · Kimi: 6
What each AI noted 8
- Metaplace 10
  $32B, no product, no revenue, no paper - the price rests entirely on who Sutskever is.
- Nemotronplace 8
  $30B valuation / $0 revenue = ∞ multiple; pre-revenue by design, valuation purely on talent bet
- Novaplace 9
  $32B valuation for pure research with no revenue or product
  Source ↗
- Claudeplace 10
  About $32B for a lab with no product and no disclosed revenue: the price rests wholly on belief.
  Source ↗
- ChatGPTplace 9
  A $32 billion disclosed financing value and a $50.4 billion model estimate coexist without a publicly established commercial revenue denominator.
  Source ↗
- StepFunplace 9
  No revenue, no filings, no problem - the whole price is one person and a thesis.
  Source ↗
- DeepSeekplace 7
  $32B, roughly 50 employees, no product and zero revenue — the purest valuation-on-promise in the table.
  Source ↗
- Kimiplace 6
  $32B for roughly 50 employees, no product and no published research - the only entry whose price rests on zero published revenue of any kind.
  Source ↗
8 AIs put it here
Source ↗
10
Google DeepMind
Alphabet files audited accounts, but none of them split out DeepMind, so it has no price of its own. The ~$100B sum-of-parts figure is one analyst estimate.
Included it: Meta, Claude, StepFun, DeepSeek, Wenxin, Qwen. Did not include: Nemotron, Nova, ChatGPT, Kimi.
In their lists: Meta: 7 · Claude: 8 · StepFun: 3 · DeepSeek: 5 · Wenxin: 9 · Qwen: 10
What each AI noted 6
- Metaplace 7
  Alphabet is fully filed, DeepMind is not - you cannot buy DeepMind's price because there is none.
  Source ↗
- Claudeplace 8
  Has no price of its own: Alphabet's public filings are the only disclosure, and they do not split out DeepMind.
  Source ↗
- StepFunplace 3
  The only top-ten name a regulator actually audits, and the only one whose AI revenue is never split out.
  Source ↗
- DeepSeekplace 5
  No standalone round and no separate revenue line: a $100B sum-of-parts estimate sits inside a $4T+ Alphabet, so the “valuation” is an analyst construction, not a traded price.
  Source ↗
- Wenxinplace 9
  No stand-alone price — Alphabet's 2.1 T cap includes DeepMind, but 30 B AI revenue gives the unit an implied ~70×
- Qwenplace 10
  An internal Alphabet engine with bundled revenue, making standalone price-to-earnings multiples impossible.
  Source ↗
6 AIs put it here
Source ↗

More about the top three ↓

02More about the top three

The ranking is above. Here is more about the three places at the top: what they are, why they are here, and what is worth knowing.

The three multiples are not like for like: Anthropic's revenue comes from an SEC filing, OpenAI's was told to investors and never filed, and Databricks' is a company-stated, unaudited run rate.

1Anthropic

No. 1 — Anthropic. On the lists of 8 of 10 AIs; median place there: 2. Its revenue sits in an SEC filing, so its roughly 14x multiple is the best backed by paper.

Did you know

"Routines" has been renamed "scheduled tasks" in the docs, but the old name survives in the web address, the CLI and the API. Because the work runs in a cloud session, it keeps going when your laptop is closed.
Source ↗

What it is: AI lab behind Claude, priced at $965B in May.
Its edge: Only one of the three whose revenue appears in a regulator filing.
Who should know: Useful for anyone who wants to see what a price looks like next to filed revenue.

Price vs published revenueabout 14x

Wikipedia · Wikidata

Why it is here

The same filing lists about $895.4 billion in compute and data-center commitments. That is a large bill next to the revenue, and it is stated in a regulator filing, which no other company in this top three has.
Source ↗

Did you know

Anthropic's docs say a cloud task can start three ways: on a schedule, from an HTTP call, or from a GitHub event. Scheduled runs are limited to 100 per hour per account, and any schedule that runs more often than hourly is rejected.
Source ↗

History

On January 12, 2026, Anthropic released Cowork as a research preview, first only for Max subscribers on macOS. Anthropic had seen people using its coding tool, Claude Code, for almost everything else, so it built Cowork as a simpler version for anyone. — neutral narrator
Source ↗

Keep in mind

The review calls usage limits the main frustration. Cowork shares a five-hour and a weekly allowance, so heavy use hits the cap faster than normal chat.
Source ↗
Scheduled Cowork tasks only run while the app is open and your computer is awake. If you close the laptop, a run waits until you reopen it.
Source ↗

What people say

the first desktop agent that's earned a place in a normal, non-technical workflow — go9x Cowork review
Source ↗

Where to go further

Read Anthropic's official docs on scheduled tasks. First look at the 'Create from the web' steps and the 'Usage and limits' table to see how many runs your plan allows.
Source ↗

Video: The Anthropic Principle - How Your Existence Could Lead to a Multiverse · Up and Atom

Compare the top three ↓

2OpenAI

No. 2 — OpenAI. On the lists of 8 of 10 AIs; median place there: 1.5. It has the biggest, freshest price, but its roughly $50B revenue was only told to investors and never filed.

Did you know

After OpenAI's October 2025 restructuring, a nonprofit, the OpenAI Foundation, controls the for-profit company. Its stake was reported at about $130 billion, close to Microsoft's roughly $135 billion stake, so a charity-style body owns a giant slice.
Source ↗

What it is: Maker of ChatGPT, priced at $1.25T.
Its edge: Largest price of the three, though not the cheapest per dollar of revenue.
Who should know: Useful for seeing how a huge price can rest on an unfiled number.

Price vs published revenueabout 25x

Wikipedia · Wikidata

History

In November 2023 OpenAI's board fired its CEO, Sam Altman. Most staff threatened to leave, and less than a week later he was back as CEO with a new board.
Source ↗

Where to go further

Open the ChatGPT plans page and compare the free plan with the paid ones. Look at what each plan includes and what it costs per month. This shows what a regular user pays for the product that the valuation is built on.
Source ↗

Nemotron, StepFun, DeepSeek, Kimi put it first. What they saw in it

Nemotron: $1.25T valuation / ~$4B ARR = 312× multiple; only ~$1.6B of ARR confirmed in company blog, rest from The Information reporting
StepFun: A $1.25 trillion price resting on a revenue number the company has never filed.
DeepSeek: $1.25T price on ~$50B of self-stated revenue is 25x — and the “markdown” from $1.4T actually raised the price per dollar of revenue.
Kimi: Wired $11B of a ~$20B round at a $1.25T post-money price - 11% below the $1.4T it first sought - while the ~$50B revenue behind its ~25x multiple is a number investors were told, never filed.

Video: Why NO ONE Plays: Sora | Super Smash Bros. Ultimate · Vars III

Compare the top three ↓

3Databricks

No. 3 — Databricks. On the lists of 8 of 10 AIs; median place there: 3. The company published its own $7B-plus run rate, which gives about 27x, but the figure is unaudited and not filed.

Agrawroh · CC BY-SA 4.0 · Wikimedia Commons

Did you know

Databricks bought the open-source AI startup MosaicML for $1.4 billion, according to Wikipedia. The Reuters article it cites says $1.3 billion, so the two sources disagree.
Source ↗

What it is: Data and AI platform company, priced at $190B.
Its edge: Publishes its own revenue figure, and it is growing over 80%.
Who should know: Useful for seeing what a price looks like when revenue is company-stated, not audited.

Price vs published revenueabout 27x

Wikipedia · Wikidata

Why it is here

The $190 billion price is about 42% above the roughly $134 billion mark from its February round. The revenue figures come from the company and are not audited public results.
Source ↗

Did you know

Databricks' open model DBRX has 136 billion parameters, but uses only about 36 billion on average to produce an answer. The article says it cost $10 million to create.
Source ↗

History

Databricks was founded in 2013 by the people behind Apache Spark, a data tool built at the University of California, Berkeley. It started as a university research project before it became a company.
Source ↗

Keep in mind

The $7 billion run rate and growth above 80% are company-reported, not audited. At $190B, that is about 27 times revenue, which leaves little room for slower growth.
Source ↗
Databricks has multibillion-dollar cloud commitments with all three big cloud providers, and its AI research is costly. Selling more shares to meet demand may upset older investors who were turned away.
Source ↗

What people say

The interest level was just insane. — Ali Ghodsi, Databricks CEO
Source ↗

Where to go further

Read the trial page first: it gives two weeks of full platform access with up to $400 in free usage. There is also a Free Edition with no expiration date.
Source ↗

Video: What is Databricks? The Story Behind the Modern Data Platform (Visual Explanation) · Data with Baraa

Compare the top three ↓

The top three side by side

	Anthropic	OpenAI	Databricks
In the AIs' lists	8 / 10	8 / 10	8 / 10
Price vs published revenue	about 14x	about 25x	about 27x
Put first by	Meta, Claude, ChatGPT	Nemotron, StepFun, DeepSeek, Kimi	—

Places with the same number of AIs are ordered by their average place in the AIs' own lists.

03Worth knowing

What the AIs noticed while building the list.

OpenAI raised $11B at $1.25T, below the $1.4T it first sought, yet the price is still about 47% above March's $852B.

Noticed by:

Included it: Meta, Claude, ChatGPT, DeepSeek, Kimi, Qwen. Did not include: Nemotron, Nova, StepFun, Wenxin.

01Final list

Anthropic

OpenAI

Databricks

Scale AI

xAI (SpaceX AI)

Perplexity

Mistral AI

Thinking Machines Lab

Safe Superintelligence (SSI)

Google DeepMind

02More about the top three

1Anthropic

Did you know

2OpenAI

Did you know

3Databricks

Did you know

The top three side by side

More ratings

03Worth knowing

04Questions people ask

05How we counted

1The question

2Why this topic

3Where it applies

4What we looked at

5Who ranked

6How the places were set

06Each AI's own list

What Meta noticed

What Nova noticed

What Claude noticed

What ChatGPT noticed

What StepFun noticed

What DeepSeek noticed

What Kimi noticed

What Qwen noticed