The short answer
Pick Gemini if you live in Google Workspace, need fast answers at low cost, or want to build apps with AI inside them through Google AI Studio. Pick ChatGPT if you want the most complete everyday chat product.
That is a split, not a knockout. Gemini won our one scored test in November 2025, and several guests have moved some daily use to it. But no one on the show has dropped ChatGPT for it, and Jordan's line has stayed the same: check the charts, then "you gotta just test it."
What we built with each
Gemini
- Solopreneur finance dashboard for Collective: Jordan started with a smaller model, but the P&L and balance sheet data was too much, so he moved to Gemini "because I was able to take the data in." It runs the cash-flow forecast and insurance compliance check (Episode 14, at 06:59).
- Mirror, Mirror on the Wall, the Guess My Costume app and the Trick-or-Treat route planner: Gemini handled the answers, costume recognition and routing, built in AI Studio (Episode 19).
- AI start-of-care packet generator: Sam built it with Gemini Pro in AI Studio in about 30 minutes (Episode 24).
ChatGPT
- Public API trading bot: the OpenAI API makes the buy and sell decisions (Episode 8).
- n8n sales outreach workflow: ChatGPT creates the structured outputs and custom emails (Episode 17).
- Ryz Score and NTRVSTA: ChatGPT helps score resumes and interviews.
Head to head in one build
The LLM Math Roaster sent the same math problems to Gemini 2.5 Pro, GPT-5, Claude Sonnet 4.5 and Grok 4 Fast, then had ChatGPT judge the Lean proofs (Episode 22).
Head-to-head
On Fermat's Little Theorem, the judge gave Gemini the top score:
So I have a leaderboard. It gave Gemini the best score. ChatGPT is 70, and so on and so forth.
— Jordan Metzner, Episode 22
Gemini scored 98. Guest Carina of Axiom spotted that one model only proved the theorem for natural numbers rather than integers, "just a close miss." One problem is not a benchmark, and Jordan said the real test would be running thousands of them.
Where each one fell short
ChatGPT. The GPT-5 launch in August 2025 annoyed Sam, whose prompts kept getting the quick answer when he wanted the thinking one (Episode 8). By December, OpenAI had declared a "code red" over Gemini's rise. Jordan's worry for OpenAI was distribution: most of what he wants from an LLM is slides, a Google Doc or a Gmail draft, and half the time in ChatGPT he is copying output into a Google Doc (Episode 23, at 21:08).
Gemini. It has not always been top tier. In Episode 27 Jordan said that in some places it "hasn't been a top tier model yet." On Gemini 3's launch day, Jordan burned all his tokens and got throttled in Cursor and Antigravity, and Sam hit "some hiccups" building a meditation app (Episode 22, at 18:50). Amar Goel of Bito gave the most balanced take:
I mean, I personally haven't felt like Gemini three is like so groundbreaking per se. I know on the benchmarks it's done really well, but I do think it's quite a good model.
— Amar Goel, Episode 23
He added that he now uses Gemini where he used to default to ChatGPT and Claude, and "sometimes I like its answers better." His bigger point was distribution: the Gemini button in Chrome, AI mode in search, and an ad model that lets Google avoid charging consumers, while no internet product has reached a billion users on a paid model.
What our guests use
- Sagi Waitzman: his team writes LinkedIn posts with Gemini, "I think for this is the best," helped by it coming with Google Workspace (Episode 28).
- Arun Kalaiselvan and Anand Chandrasekaran, Arya Health: Gemini through Vertex AI in production, which helps with HIPAA-compliant workflows.
- August Kiles, Emblem: uses Claude and Gemini the most in the product, plus OpenAI and Grok.
- Tristan Wilson: construction executives mostly use ChatGPT, and OpenAI enterprise accounts are spreading.
- Ben Lerner, who once worked on language processing at Google: "all of these models are keeping pace with each other."
Which one we reach for now
- July 2025: Jordan said OpenAI was "still the leader," with Google catching up (Episode 5).
- November 2025: Gemini 3 impressed Jordan "for coding and design taste."
- January 2026: Jordan said we would probably move to Gemini if it became "top notch" (Episode 29).
- May 2026: about to try Gemini 3.5 Flash in Antigravity, Jordan said Google's TPUs and low token costs "could provide for a pretty dangerous recipe" with a really good model (Episode 43).
Today we use Gemini mostly inside Google AI Studio for fast prototypes, and ChatGPT for quick questions and PRDs. For coding and analysis we lean on Anthropic and OpenAI's coding tools; see Claude vs Gemini and Claude vs ChatGPT. For images, Google's Nano Banana is its own story in ChatGPT vs Nano Banana.
FAQ
Is Gemini better than ChatGPT?
In our LLM Math Roaster, Gemini 2.5 Pro scored 98 on a Fermat's Little Theorem proof while GPT-5 scored 70, with ChatGPT as the judge. Guest Amar Goel found Gemini 3 quite good but not groundbreaking, and said the ChatGPT product is richer.
Is Gemini cheaper than ChatGPT?
We did not compare prices ourselves. Carina of Axiom said Gemini 3 is much cheaper to run than GPT, and David Petrou called Gemini 3.8 Flash strong on cost versus quality.
Is Gemini faster than ChatGPT?
Anand Chandrasekaran of Arya Health said Gemini 3 and Gemini 2.5 Flash respond so fast they are just too good. We never timed the two side by side.
Which is better for coding, Gemini or ChatGPT?
When Gemini 3 launched, Jordan was impressed with its coding and design taste. For day-to-day coding he has favored OpenAI's Codex, so neither chat app is where we do most coding.