Gemini vs ChatGPT: Gemini 4 Argon vs GPT-6 Astra on the Same 4 Builds

ChatGPT's top model, GPT-6 Astra, beat Google's Gemini 4 Argon in all four builds when Pat Simmons gave both the same hard jobs in October 2026: a photoreal 3D product bottle, a product launch video, a car soccer game and a rebuild of a 3D-heavy product website. Argon came last of the four models in every round.

Argon was far cheaper to run, though. All four of its builds came to $39.54, against $409.96 for Astra, and most of Astra's bill came from one build, the website rebuild, which ran for 499 minutes and cost $357.84. The full results, prompts and live builds are in the full Gemini 4 Argon review and test results.

Which one you can use also depends on your plan. In the Gemini app, Argon comes with Google AI Pro at $19.99 a month or Ultra. In ChatGPT, Plus ($20 a month) includes Astra in Work and Codex, and Pro's top reasoning option in chat runs on it, per OpenAI's help page on Astra (checked November 5, 2026).

Ranks are out of four models (Opus 5.5, Fable 5.1, GPT-6 Astra and Gemini 4 Argon), from Pat's blind picks in his October 2026 Argon review video at the top of this page. Costs, tokens and times are from the cost tables in the Argon review. Opus 5.5 won all four rounds.

Is Gemini better than ChatGPT in 2026?

For long, hard build work, ChatGPT's top model did better. GPT-6 Astra placed ahead of Gemini 4 Argon on every one of the four builds Pat ranked blind, and Argon placed last in all four, behind Anthropic's Opus 5.5 and Fable 5.1 as well.

Pat's own verdict at the end of the review was: "So for me, it still makes sense to use the Claude models or Astra."

That verdict covers build work only. Pat didn't compare everyday chat, writing, research, images or voice in Gemini and ChatGPT, so this test can't tell you which app answers questions better. He said as much in the video, calling them "very limited tests" and adding that Argon "could be in more day-to-day tasks much better of a model."

The 4 builds: who won each one

Opus 5.5 won every round. Between the two models in this comparison, Astra finished ahead of Argon each time, coming third on the 3D bottle and the car soccer game and second on the launch video and the website rebuild.

A photoreal 3D cold brew bottle you can orbit, open and pour

Each model had to build a 3D bottle for a made-up coffee brand that you can spin around, crack open and pour into a glass of ice, in the browser and photoreal. Astra came third and Argon fourth. Pat said the two were "relatively close" here, and that "Astra did just a little bit better" while "Argon just got very lazy in the actual cup."

A 30 to 60 second product launch video

The second job was a launch video for the same coffee brand, with photoreal product shots, music and sound. Astra came second and Argon fourth. Pat said Argon's video "did not feel like a proper, you know, commercial or, you know, social video," though its 3D work was better than on the bottle.

A playable car soccer game in the style of Rocket League

Next, each model built a one-on-one car soccer game against a computer opponent, with photoreal graphics, in the browser. Astra came third and Argon fourth. On the reveal, Pat said of Argon's game: "That's just not a good generation."

A rebuild of the oryzo.ai website, 3D and motion included

The last job was to rebuild the oryzo.ai website pixel for pixel, making every image and 3D model from scratch. Astra came second and Argon fourth. Astra spent 499 minutes on it, which Pat called "literally 8 hours of nonstop work in Blender," and he said it showed in the result, since Astra got the site's hand interaction right.

Which one is cheaper to run?

Gemini 4 Argon was cheaper on every build, $39.54 for all four against GPT-6 Astra's $409.96. Leave out the website rebuild, where Astra ran for over eight hours, and the gap shrinks to about $52 for Astra against about $28 for Argon across the other three builds.

These are API-equivalent costs: each build's real token count (tokens are the small chunks of text a model reads and writes) priced at the model's published rates. Pat ran the builds on subscriptions, and on Google's $99-a-month plan he hit the daily usage limit after about one build, the 3D bottle.

Argon's low bill comes partly from a sale. Google's API pricing page lists Argon at $2 per million input tokens and $10 per million output through November 30, 2026, then $4 and $20 from December 1. OpenAI's API pricing page lists GPT-6 Astra at $10 per million input tokens and $50 per million output (both checked November 5, 2026).

The price per token only tells part of the story, because the two models used tokens very differently. Argon wrote far more output than Astra on three of the four builds, while Astra's website rebuild alone read 271.2 million tokens of input. Pat made the same point in his GPT-6 Astra launch video, comparing Astra with Fable 5.1: "It's less about the input and output costs and more about how efficiently it actually takes a model to get a task done."

Which plan gets you each model?

Gemini 4 Argon comes with Google AI Pro ($19.99 a month) and Ultra ($99.99 or $199.99 a month) in the Gemini app. GPT-6 Astra is in ChatGPT Work and Codex on Plus ($20 a month), and powers the top reasoning option in chat on Pro, which starts at $100 a month. Plan details for Google are from Google's subscriptions page and for OpenAI from ChatGPT's pricing page and OpenAI's help center, all checked November 5, 2026.

Here is where each row comes from. Argon on AI Pro and Ultra, the Plus and Pro prices and Ultra's 5 and 20 times usage tiers are on Google's subscriptions page. Google's support page says that for people without a plan, "Most prompts will go to Flash-Lite, our fastest model." Argon in Antigravity needing Ultra is from Google's launch post, as covered in how to get Gemini 4 Argon in the Gemini app.

On the ChatGPT side, Luna on Free and Go is from ChatGPT's pricing page, the $20 Plus price from OpenAI's Plus help page, and the three Pro prices from OpenAI's Pro tiers page. OpenAI's help page on Astra says "Plus includes Astra in Work and Codex, but not GPT-6 Pro in Chat," and that GPT-6 Pro in chat "is powered by Astra" on eligible Pro, Business and Enterprise plans. Sol in Plus chat is covered in which GPT-6 model you get on each ChatGPT plan.

From December, ChatGPT Plus, Pro and Business share one usage allowance across ChatGPT and Codex, so on Plus, long Astra jobs in Work or Codex draw from the same allowance as your chats. That change is explained in how ChatGPT's shared usage limit works.

What is the downside of Google Gemini?

In Pat's tests, the biggest downside of Gemini 4 Argon was the quality of what it built: it came last on all four builds, and its 3D work was the weakest of the four models. On the website rebuild Pat said, "It's clearly worse at 3D than the other models," and when Argon's bottle was revealed he asked, "Hey Google, what happened to your pagoda here?"

The review and Google's own pages turn up three more problems:

• The usage limit ran out fast: Pat hit the daily cap on the $99 plan after about one build.
• The cheap API price ends on November 30, when Argon goes to $4 and $20 per million tokens, per Google's pricing page.
• Argon is missing from Google's free and $4.99 Plus plans, and coding with it in Antigravity needs Ultra.

Astra's bill is harder to predict. On the website rebuild it kept working for 499 minutes and ran up $357.84, and Pat said of Astra on long jobs that "sometimes it'll just be a workhorse and then other times it'll just stop and think it's good enough."

Can Gemini do everything ChatGPT can?

For everyday office work the two apps now overlap a lot, though Pat's test only compared them on builds. Gemini can now edit the PDFs, Docs, Sheets and slide decks it makes right in the chat, and Gemini in Google Slides can rework a whole deck from one prompt.

ChatGPT, for its part, can read and act in your Chrome tabs through its extension, connects to iCloud Mail and Calendar on every plan, and has Slides your team can edit together.

Pat's take in the review was that Gemini's Google Docs advantage has mostly gone, because ChatGPT and Claude now have connectors that reach Google Docs, Sheets and Slides. Claude's version is covered in Claude now works inside Google Docs, Sheets and Slides. For a wider look at the apps that compete with ChatGPT and what each costs, see ChatGPT alternatives, and for the company behind it, OpenAI explained.

How Pat tested them

Every model got the exact same prompt for each build, with no steering, and ran at high effort. Pat judged each round blind, with the builds labeled A to D, and only then revealed which model made which. The four models were Opus 5.5, Fable 5.1, GPT-6 Astra and Gemini 4 Argon.

The models worked as coding agents, writing code and driving Blender (the free 3D program) with scripts. They ran on subscriptions, and the costs shown are API-equivalent: each build's real token counts priced at the model's published per-token rates. Every prompt, every live build and the cost table for all four models are in the full Gemini 4 Argon review.

Questions people ask

Is Gemini 4 Argon free?

No. In the Gemini app it needs Google AI Pro ($19.99 a month) or Ultra, per Google's subscriptions page, and the API has no free tier for Argon, per Google's API pricing page.

Is GPT-6 Astra free?

No. ChatGPT's pricing page shows no Astra on Free or Go. Plus ($20 a month) includes it in Work and Codex, and Pro uses it for top reasoning in chat, per OpenAI's help page.

Google AI Pro or ChatGPT Plus: which gets you more?

The two cost about the same, $19.99 and $20 a month. AI Pro gets you Gemini 4 Argon in the Gemini app, and Plus gets you GPT-6 Sol in chat plus a limited amount of GPT-6 Astra in Work and Codex. In Pat's builds, Astra placed ahead of Argon every time.

Gemini vs ChatGPT for coding: which did better?

All four builds were coding jobs, and GPT-6 Astra placed ahead of Gemini 4 Argon on each one. To code with Argon in Google's Antigravity agent you need Ultra, while Plus includes Astra in Codex, OpenAI's coding agent.

What is the strongest version of Gemini?

Gemini 4 Argon. Google's API pricing page calls it "Our flagship reasoning model," and the Gemini app labels it for "Complex tasks."

Will Gemini 4 Argon get more expensive?

Yes, on the API. Google lists $2 per million input tokens and $10 per million output through November 30, 2026, then $4 and $20 from December 1. That is still below GPT-6 Astra's $10 and $50.

Is Gemini better than ChatGPT for writing or research?

This test can't say. Pat compared the two models on four build jobs and didn't test writing, research, images or voice.

Update November 2026

November 5, 2026: First published. Prices and plans checked on Google's and OpenAI's pages that day. Gemini 4 Argon's sale price on the API ends on November 30, so this page will be updated after that.

Sources

• AI for Mortals: Gemini 4 Argon review, test results and costs
• AI for Mortals: GPT-6 in ChatGPT, which model you get on every plan
• Google: Gemini subscription plans (checked November 5, 2026)
• Google: Changes to Gemini model access and limits
• Google AI for Developers: Gemini API pricing (checked November 5, 2026)
• OpenAI: API pricing (checked November 5, 2026)
• OpenAI: ChatGPT pricing (checked November 5, 2026)
• OpenAI Help: What is ChatGPT Plus
• OpenAI Help: About ChatGPT Pro tiers
• OpenAI Help: Managing usage with GPT-6 Astra in Work and Codex