Gemini Team Answers ‘Gemini Is Cooked’ as Neon Ranks #2

Google’s Gemini 4 Neon ranked #2 overall on Arena’s Agent Arena Index with 1,336 points. In the same week, Google DeepMind staff publicly answered SemiAnalysis’s August call that “Gemini is Cooked”.

Demis Hassabis wrote on X early on Friday, replying to an apology post he did not name: “Apology accepted”, adding that it had missed checking the “Underestimated the cracked Google DeepMind team” box.

Important: not investment advice

This article is for information and education only. It is not investment, financial, legal or tax advice, and it is not a recommendation or an offer to buy or sell any security, ETF, option or other financial product. FourWeekMBA is not a broker-dealer or a registered investment adviser. IPO shares, leveraged and inverse ETFs, and options carry significant risk, including the loss of your entire investment; leveraged and inverse ETFs reset daily and can lose value quickly. Prices, availability, fees and eligibility change and can differ by country and broker, so check the prospectus and your broker’s terms. Figures marked “our arithmetic” are our own calculations from the cited filings and may contain errors. Do your own research and consider speaking to a licensed financial professional before making any investment decision.

Business Pill · The Model and the Harness

An evergreen Business Engineer explainer on models and the harnesses around them. It teaches the general idea only and does not discuss Gemini 4 or Arena.

The key insight: Arena ranked Gemini 4 Neon #2 on its Agent Arena Index with 1,336 points, the top Google model; in the same week, Google DeepMind staff publicly answered SemiAnalysis’s August “Gemini is Cooked” call; Google Cloud grew about 102.6% in the third quarter (our arithmetic).

What Arena Measured

Arena said Neon (High) is the best Google DeepMind model on its Agent Arena Index, based on 4,200 real-world agentic sessions from its community.

It placed Neon above three sibling models: Carbon (High) at #4 with 1,303 points, Hydrogen (High) at #9 with 1,271 and Argon (High) at #11 with 1,229.

Arena put Neon’s median cost at $2.82 per task. It said Neon scores higher than #3 Claude Opus 5.5 (Max) for 53% less per task, and costs 69% less per task than #1 Claude Fable 5.2 (Max).

On Arena’s sub-signals, Neon ranked #2 on Praise vs. Complaint (1,539 points), #2 on Steerability (1,298) and #3 on Task Completion (1,316).

Arena’s scores for Google’s four Gemini 4 models; Neon is 33 points ahead of Carbon (our arithmeti
Arena’s scores for Google’s four Gemini 4 models; Neon is 33 points ahead of Carbon (our arithmetic).

What Was Posted on X

Arthur Douillard of Google’s Gemini RL scaling team, 19 Nov, quoting SemiAnalysis’s August report ‘Gemini is Cooked but GCP is Cooking’ and saying the team printed it and kept it on their desks.

Demis Hassabis, 20 Nov, replying ‘Apology accepted’ to an apology post he does not name, and praising the Gemini and Google DeepMind teams.

Wyatt Walls, an X user, 19 Nov: ‘Gemini 4 Neon has the best CoT’. One user’s assessment, not a benchmark.

The Number

Neon’s 1,336 points sit 33 above Carbon and 107 above Argon, the first Gemini 4 model (our arithmetic on Arena’s scores).

Google DeepMind announced the rollout on 18 November, saying Neon sits “between Carbon and Hydrogen”. We reported its $3 and $15 per-million-token pricing the same day.

The SemiAnalysis Call

On 7 August, SemiAnalysis posted its report “Gemini is Cooked but GCP is Cooking”, summarised as “DeepMind’s long term failure is Google Cloud’s short term gain”.

On 19 November, Arthur Douillard of Google’s Gemini RL scaling team wrote that the team “printed & put that article on our desks to be reminded daily of what used to be true at the time”. His post drew about 391,000 views by Friday.

SemiAnalysis’s Dylan Patel had already posted on 2 November: “Official apology to Google from me, they cooked with Carbon”, calling the model “spikier in terms of domains and capabilities”.

Half the Thesis Held

The cloud half of the call has held so far. Alphabet’s third-quarter 10-Q shows Google Cloud revenue of $30.709 billion, up from $15.157 billion a year earlier.

That is growth of about 102.6%, and Cloud’s operating income of $11.121 billion is a 36.2% margin (our arithmetic on the 10-Q). Cloud was about 24.1% of Alphabet’s $127.577 billion quarterly revenue.

Bar chart: Google Cloud revenue rose from $15,157 million in Q3 2025 to $30,709 million in Q3 2026
Google Cloud’s third-quarter revenue from Alphabet’s 10-Q, about 102.6% growth (our arithmetic).

Wyatt Walls, an X user, posted that “Gemini 4 Neon has the best CoT”, meaning chain of thought. That is one user’s assessment, not a benchmark.

The Structural Read

As we read it, the August call split into two bets: Google Cloud has kept its side, while Arena’s agent ranking is the strongest public data point this week on the model side.

As we read the leaderboard, Arena’s cost comparison is the sharper claim: a #2 score at $2.82 a task puts price, not just rank, in the contest.

Arthur Douillard, Gemini RL scaling team

“we printed & put that article on our desks to be reminded daily of what used to be true at the time”

Three Implications

FOR MODEL BUYERS Arena puts Neon’s median cost at $2.82 per task, 53% below Claude Opus 5.5 (Max) by its figures.

FOR GOOGLE Neon is Google’s highest-ranked model on Arena’s agent index, at #2 overall.

FOR THE CLOUD THESIS Alphabet’s 10-Q shows Google Cloud revenue up about 102.6% in the third quarter (our arithmetic).

What Is Not Established

Arena’s index reflects its community’s sessions and votes; other leaderboards may rank the models differently. We have not run our own tests.

Hassabis did not name the post he was answering, and his reply asked whether Anthropic’s Sholto Douglas was behind it. We could not confirm which apology post he meant.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The AI Supercycle

The Business Engineer map of who controls the bottlenecks in the AI stack.

Read the Map of AI →

The Bottom Line

Arena ranked Gemini 4 Neon #2 on its Agent Arena Index with 1,336 points. In the same week, Google DeepMind staff, including Hassabis, publicly answered SemiAnalysis’s August “Gemini is Cooked” call. Google Cloud’s 102.6% third-quarter growth matches the report’s other half.

Important: not investment advice

This article is for information and education only. It is not investment, financial, legal or tax advice, and it is not a recommendation or an offer to buy or sell any security, ETF, option or other financial product. FourWeekMBA is not a broker-dealer or a registered investment adviser. IPO shares, leveraged and inverse ETFs, and options carry significant risk, including the loss of your entire investment; leveraged and inverse ETFs reset daily and can lose value quickly. Prices, availability, fees and eligibility change and can differ by country and broker, so check the prospectus and your broker’s terms. Figures marked “our arithmetic” are our own calculations from the cited filings and may contain errors. Do your own research and consider speaking to a licensed financial professional before making any investment decision.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Ranks, scores, session count and cost comparisons are Arena’s own figures from its X post; the rollout line is Google DeepMind’s. Quotes are from the named X posts. Google Cloud figures are from Alphabet’s third-quarter Form 10-Q. The 33- and 107-point gaps, 102.6% growth, 36.2% margin and 24.1% share are our arithmetic.

Sources: Arena on X, 18 Nov 2026: Gemini 4 Neon (High) #2 on the Agent Arena Index · Google DeepMind on X, 18 Nov 2026: Gemini 4 Neon rollout · Demis Hassabis on X, 20 Nov 2026 00:03 UTC · Arthur Douillard (Google DeepMind) on X, 19 Nov 2026 · Dylan Patel (SemiAnalysis) on X, 2 Nov 2026 · SemiAnalysis on X, 7 Aug 2026: Gemini is Cooked but GCP is Cooking · Wyatt Walls on X, 19 Nov 2026 · Alphabet Inc. Form 10-Q for the quarter ended 30 Sep 2026, segment revenue and operating income

Scroll to Top
FourWeekMBA