Skip to content
Gemini 4 Neon vs Opus 5.5 and GPT-6.2 Sol: Which One for Everyday Work?
Model comparisons

Gemini 4 Neon vs Opus 5.5 and GPT-6.2 Sol: Which One for Everyday Work?

Gemini 4 Neon, Opus 5.5 and GPT-6.2 Sol all got our five office jobs right, so the choice comes down to speed, price and style.

AI for Mortals Staff

10 min read

0

If you pay for one AI plan and wonder whether another one is better, the Gemini Pro vs ChatGPT Plus vs Claude Pro question just changed. On November 18, Google added Gemini 4 Neon to the Gemini app for Google AI Pro and Ultra subscribers, and it's now the strongest model a $19.99 Pro plan gets.

Google compares Neon with two rivals by name: Claude Opus 5.5, the top model on Claude Pro, and GPT-6.2 Sol, OpenAI's everyday model on ChatGPT Plus. All three plans cost about $20 a month.

So we gave Neon and Opus 5.5 the five office jobs we ran on GPT-6.2 Sol when it came out: a messy email thread, a 48-page report, a contact list full of duplicates, a 10-slide deck and a small budget tool. Each model ran every job twice, and we checked the results against an answer key.

All three got all five jobs right. The differences were in speed, price and how each one handled the gray areas.

Gemini Pro vs ChatGPT Plus vs Claude Pro: the short answer

  • If you already pay for one of them, stay put. On everyday office work, none of the three got anything wrong that the others got right.
  • Gemini 4 Neon writes the most complete drafts and was quickest on the email and the calculator, but it used the most tokens. If you already pay for Google AI Pro, make it your default.
  • Claude Opus 5.5 checks the details around the job. It flagged things to check before you hit send, and in one run it left out a fee the client hadn't agreed to yet.
  • GPT-6.2 Sol writes the shortest answers and costs the least per job by a wide margin, so it's the one to pick if you live on the API or keep running into plan limits.
  • For decks, Sol made the slides you'd read from the back of the room. Neon and Opus made polished dashboards with a lot of small print.

Which model each $20 plan gives you

Here's where each model lives, with the price of the main plan and the developer price per million tokens in and out:

  • Gemini 4 Neon: Google AI Pro ($19.99 a month) and Ultra. Pick 4 Neon, labeled "Deeper analysis," from the model menu. API: $3 in, $15 out.
  • Claude Opus 5.5: Claude Pro ($20 a month, or $17 billed yearly) and up. Free accounts don't get Opus. API: $4 in, $20 out.
  • GPT-6.2 Sol: ChatGPT Plus ($20 a month) and up. API: $2 in, $10 out.

Neon costs a quarter less than Opus per token and half again as much as Sol. Our Gemini pricing guide and Claude pricing guide break down what else each plan includes.

The token counts further down are a rough guide to which model stretches a plan furthest.

What Google's own tests say about Neon vs Opus 5.5 and Sol

Google's launch table puts Neon against Opus 5.5 and GPT-6.2 Sol on 19 tests. Neon scores highest on 16 of them, with its biggest leads on office automation, very long documents and security work.

Google's benchmark table comparing Gemini 4 Neon, Claude Opus 5.5 and GPT-6.2 Sol across 19 tests in capability indices, agentic coding, knowledge work, science and math, reasoning, multimodal understanding, long context and cybersecurity. Neon's column is highlighted as best on most rows; Opus 5.5 leads on Terminal-bench 4.1 and FrontierCode 2.0, and GPT-6.2 Sol leads on Chartography.
Google's comparison table for Gemini 4 Neon. Opus 5.5 leads on two coding tests and GPT-6.2 Sol on reading charts. Source: Google

Opus 5.5 still wins two of the coding tests by about a point, and Sol reads dense business charts more accurately than Neon. These are Google's numbers, run at settings Google chose.

On the independent Artificial Analysis leaderboard, Neon scores 61 at its highest setting, Opus 5.5 59 and Sol 57.

Google's cost chart tells the other half of the story. Neon is cheaper per task than Opus 5.5 at every setting, but GPT-6.2 Sol does its best work for about a dollar a task, a fraction of either.

Google's chart of Artificial Analysis Intelligence Index score against cost per task in dollars. GPT-6.2 Sol's line sits far left, reaching 57 for about one dollar. Gemini 4 Neon climbs from 44 at Low to 61 at Xhigh for about six dollars. Gemini 4 Carbon reaches 60 for under eight dollars, and Claude Opus 5.5 reaches 59 at about ten dollars.
Score against cost per task for each thinking level. Up and to the left is better. Source: Google

Five everyday jobs we gave Gemini 4 Neon, Opus 5.5 and Sol

Neon ran through the Gemini API at its default high thinking level. Opus 5.5 ran in Claude Code at high effort, and GPT-6.2 Sol in Codex at high reasoning, using its results from our November 17 test with the same prompts.

Claude Code and Codex can write and run small bits of code while they work, which the Claude and ChatGPT apps can do too. Neon through the API couldn't, so it had to do everything in its head. The prompts are below so you can paste them into whichever app you pay for.

A messy email thread: Opus caught what the others let through

Six emails between a clinic and a workshop trainer, with the date, start time, budget and travel cap all changing along the way, plus a question about laptops and a request to keep patient data out.

Reply to a messy email thread
Here's a messy email thread. Write the reply Dana asked for, addressed to everyone on the thread. Keep it short and friendly, and include a short agenda for the workshop.

[paste the whole thread here]

All three confirmed Friday, November 20, 8:30 to noon, in the clinic's conference room for 14 people. All three told Priya laptops were optional and promised no patient data.

Neon wrote the longest reply, with a timed agenda that wraps up at 11:50 so the front desk is ready when the clinic opens at noon. GPT-6.2 Sol wrote the shortest, about a third of Neon's length, but its agenda had no times.

Opus 5.5 did something the other two didn't. Both of its runs ended with a short "before you send" note. In one, it pointed out that Dana's deadline for this reply had already passed. In the other, it left the fee out entirely. The trainer had quoted "$2,800 plus travel," and the client offered $2,800 with travel capped at $150, which the trainer never agreed to. Opus suggested settling that in a separate email before confirming it to everyone.

It also slipped once. One Opus subject line read "confirmed for Thu Nov 20," though November 20 is a Friday, which its own email body got right. Read the subject line too before you send.

A 48-page report: five for five on all three

We pasted in NIST's AI Risk Management Framework and asked for a 150-word summary plus five questions. One is a trap: it asks what penalties the framework sets, and it sets none.

Summarize a long document and answer questions
Below is a long document. First, summarize it in about 150 words for a busy manager. Then answer these five questions, quoting the document where you can:
1. What are the framework's core functions, and which one is described as cross-cutting?
2. List the characteristics of trustworthy AI the framework names.
3. How many categories does each core function have?
4. Is the framework mandatory, and what penalties does it set for organizations that don't follow it?
5. What does MANAGE 4.1 ask organizations to do?

<document>
[paste the document here]
</document>

All three answered every question correctly and caught the trap, in every run. We matched every quote against the report, and none was made up.

Neon and Opus quoted the report two to three times as often as Sol, which is handy when you need to show your boss where an answer came from. Neon also wrote the longest answers, about 600 words against roughly 500 for Opus and 360 for Sol.

A contact list full of duplicates: Neon did it without running any code

A 172-row contact list with duplicates, blank rows, mixed date formats, messy phone numbers, and six pairs of different people who share a name.

Clean up a contact list
Clean up this contact list. Two rows are the same person only when the email address matches (ignoring capitals and stray spaces); two people can share a name. Drop blank rows. Merge duplicate people into one row each, and when the copies disagree on the signup date, keep the earliest. Format names as First Last in normal capitalization, emails in lowercase with no spaces, phone numbers as (555) 555-5555, states as two-letter codes, and dates as YYYY-MM-DD (dates with slashes are month/day/year). Return the cleaned CSV in a code block, then say how many rows are left, how many duplicate rows you removed, and list any people who share a name but are different people.

[paste the CSV here]

We checked every row against the right answer. All three returned exactly 146 people with zero mistakes, removed the 26 duplicates and three blank rows, and named all six pairs of people who share a name.

Opus and Sol each wrote a short script to clean the list. Neon had no way to run one through the API, so it reasoned through all 172 rows and still made no mistakes. That took it about four minutes, the same as Sol and twice as long as Opus, and most of Neon's thinking tokens for the whole test went into this one job.

A 10-slide deck: three very different looks

One page of rough notes from a support team's quarter, turned into a 10-slide all-hands deck.

Turn notes into a 10-slide deck
Turn these notes into a 10-slide presentation for a Monday all-hands. Make it a single self-contained HTML file: each slide is a 16:9 section (1280x720) stacked vertically, clean modern design, readable from the back of a room, no external libraries or images. Use only the numbers in the notes. Return only the HTML in one code block.

[paste your notes here]

All three decks had 10 slides, and every number on them came from the notes.

Three rows of four slides each, made from the same notes. Top row, Gemini 4 Neon: a navy title slide, then white dashboard-style slides with cards, progress bars and small labels, showing 18,420 tickets, 4h 10m response time, a 312 backlog and an ask for a $4k a month contractor. Middle row, Claude Opus 5.5: a dark title slide reading Q3 Review, then cream slides with colored cards, big 6h 45m to 4h 10m numbers, and a backlog slide with 312 in large red type. Bottom row, GPT-6.2 Sol: purple and cream poster-style slides with heavy black headlines, big numbers and very little text.
Four slides from each deck: the title, the response-time slide, the problem slide and the request to leadership. Top: Gemini 4 Neon. Middle: Claude Opus 5.5. Bottom: GPT-6.2 Sol. Our test.

Neon and Opus both built polished, dashboard-style decks with cards, progress bars and color-coded labels. They look great on a laptop, but both packed in small print that won't read from the back of a room, which the prompt asked for.

Opus wrote the stronger headlines, like "We're answering customers much faster," and a timeline slide for the self-serve password reset. Neon's slides are the busiest of the three, with small explanation paragraphs on most of them.

GPT-6.2 Sol made the slides you could actually present: a few words, huge numbers, bold purple headlines. If you make slides with AI, ask for "one point and one number per slide" whichever model you use.

A budget calculator: all three passed every check

A one-page budget tool: take-home pay, expenses by category, a savings goal, and a warning when you're short.

Build a budget calculator
Build a monthly budget calculator as a single self-contained HTML file (no external libraries). I want to enter my monthly take-home pay, add and remove expenses (name, amount, and a category picked from Housing, Food, Transport, Bills, Fun, Other), and set a savings goal as a percent of my pay. Show total spending, what's left over, how much of my pay I'm actually saving as a percent, a per-category breakdown, and a clear warning when I'm short of my savings goal or spending more than I earn. Everything updates as I type. Keep my budget if I close the tab and come back. Handle bad input (empty, negative, letters) gracefully: treat it as 0 and don't show NaN. Make it look good on a phone and a laptop. Return only the HTML in one code block.

We ran each calculator through 12 automatic checks: the math, the warnings, overspending, letters typed into a money box, removing a row, closing and reopening the tab, and fitting on a phone. All three passed all 12, in every run.

Three budget calculators side by side, each showing take-home pay of $4,000, expenses for Rent $1,500, Groceries $400 and Car $500, total spending of $2,400.00, $1,600.00 left over and a 40% savings rate. Left, Gemini 4 Neon: white cards with a blue Add Expense button and a category breakdown with colored bars. Middle, Claude Opus 5.5: a light gray layout with a summary panel and a green 'You're on track for your savings goal' message. Right, GPT-6.2 Sol: a cream page with a large serif Monthly Budget title and a dark green summary box.
The three budget calculators with the same numbers typed in. Left to right: Gemini 4 Neon, Claude Opus 5.5 and GPT-6.2 Sol. Our test.

Neon built its calculator fastest, in about a minute and a half on average. Opus took about three and a half minutes, because it tested its own page before answering.

Speed and cost: Gemini 4 Neon vs Opus 5.5 vs GPT-6.2 Sol

Here's the whole five-job set added up. Times and costs are averages per run. Cost is what the tokens cost at API prices; on a $20 plan you pay the monthly fee instead, and the token count is roughly what counts against your limit.

  • Gemini 4 Neon: 5 of 5 right in both runs, 8 min 43 s, about 78,000 tokens (including thinking), $1.27 a run
  • Claude Opus 5.5: 5 of 5 right in both runs, 8 min 16 s, about 63,000 tokens, $2.88 a run
  • GPT-6.2 Sol: 5 of 5 right (one run), 9 min 15 s, about 26,000 tokens, $0.51 a run

The three finished within a minute of each other. Neon was quickest on the email and the calculator, Opus on the report and the spreadsheet, and Sol on the deck.

The real gap is in tokens. Neon used about three times as many as Sol, mostly thinking its way through the spreadsheet, and Sol did the same work for a fraction of the price, which matches Google's own cost chart above.

Opus costs the most per run partly because Claude Code sends its own instructions and tools with every job. In the Claude app, a job like these counts against your Pro limit rather than costing money.

What people are saying about Gemini 4 Neon

The loudest reaction to the launch was about the names. In the biggest Reddit and Hacker News threads, a large share of the top comments asked which of Argon, Carbon, Hydrogen and Neon to use, and whether Neon replaces Argon.

Among people who have tried it, the first thing they mention is speed, with near-Carbon answers at Hydrogen-like pace. On Hacker News, a lot of the talk was about whether it can stand in for Opus 5.5 for coding, since it costs less per token.

Which AI is best for everyday work: Neon, Opus or Sol

Pick Gemini 4 Neon if you already have Google AI Pro, through a Google One plan or a student offer, or you live in Gmail and Docs. It matched the other two on every job here and wrote the fullest drafts, and Google says it pulls details from Gmail and Docs into its answers.

Pick Claude Opus 5.5 for work where a wrong detail costs you, like a client email or a contract summary. It was the only one that stopped to flag a fee nobody had agreed to and a deadline that had already passed.

Pick GPT-6.2 Sol if you want short answers, plan to build on the API, or keep hitting your plan's limit. It got everything right for a fraction of the cost.

Paying for a second plan doesn't make sense for this kind of work. All three get emails, documents, spreadsheets, slides and small tools right, so use the one you have and spend the $20 elsewhere. If you're choosing your first plan, start with where your files already live.

For harder builds, Pat put Opus 5.5 against Gemini 4 Argon and GPT-6 Astra in his Argon review, and our Claude vs ChatGPT vs Gemini build test has the full results.

For more on Neon itself, including how it compares with Carbon and Hydrogen inside Gemini, see our Gemini 4 Neon launch explainer.

Sources

What did you think of this post?

Comments

Sign in to join the conversation.

You might also like.

Articles, builds, skills and tools on the same idea.
More Model Comparisons

AI FOR MORTALS.

Your daily AI edge that will make everyone wonder how you do it all.

Every weekday, the practical AI updates that matter, and exactly what to do with them.

The latest in AI, in plain English. Free, every weekday.