Gemini vs ChatGPT

What the names mean

Gemini is Google's chat app and the name of its family of models. ChatGPT is OpenAI's chat app, built on OpenAI's GPT models.

We tested the models themselves through OpenRouter, and whichai.chat sends your messages the same way. Features of the Gemini and ChatGPT apps aren't part of the comparison.

How Gemini and GPT models scored on harder prompts

The average score across 36 harder prompts asked twice, so 72 answers per model.

Scores on 36 harder prompts, 8 November 2026

Model	Score	Full marks	Typical reply time	Cost of all answers
GPT-6.1 Sol OpenAI	99.7%	71 of 72	4.8 s	$0.234
GPT-6 Astra OpenAI	99.4%	70 of 72	4.3 s	$1.15
Gemini 3.8 Flash Google	98.9%	70 of 72	12.1 s	$0.823
Gemini 4 Argon Google	98.2%	68 of 72	5.7 s	$0.293
GPT-6 Luna OpenAI	95.6%	68 of 72	3.3 s	$0.020

• GPT-6.1 Sol OpenAI 99.7%

  Full marks

  71 of 72

  Reply time

  4.8 s

  Cost

  $0.234

• GPT-6 Astra OpenAI 99.4%

  Full marks

  70 of 72

  Reply time

  4.3 s

  Cost

  $1.15

• Gemini 3.8 Flash Google 98.9%

  Full marks

  70 of 72

  Reply time

  12.1 s

  Cost

  $0.823

• Gemini 4 Argon Google 98.2%

  Full marks

  68 of 72

  Reply time

  5.7 s

  Cost

  $0.293

• GPT-6 Luna OpenAI 95.6%

  Full marks

  68 of 72

  Reply time

  3.3 s

  Cost

  $0.020

Scored with Raq.com's AI Benchmarker on 8 November 2026. One answer is worth about 1.4 points, so GPT-6.1 Sol and Gemini 3.8 Flash, the best of each, are less than one answer apart.

What we found

One prompt did most of the damage: a decision record written from an email thread, where models left out who had agreed to do what. Both Gemini models lost 0.4 or more each time. The GPT models lost 0.2 at most.

Gemini 4 Argon also went one word over a 90-word limit and got one figure wrong on a timesheet. GPT-6 Luna, the cheapest of the five at about 2 cents for all 72 answers, got the statutory redundancy pay question wrong on both tries.

The GPT models gave the quicker replies, from 3.3 to 4.8 seconds, against 5.7 for Gemini 4 Argon and 12.1 for Gemini 3.8 Flash, which despite the name was the slower of the two.

Where the best of each lost marks

GPT-6.1 Sol and Gemini 3.8 Flash scored highest for their companies. These are the prompts where either dropped points, with the reason.

Prompts where Gemini 3.8 Flash or GPT-6.1 Sol lost marks

Prompt	Gemini 3.8 Flash	GPT-6.1 Sol
Decision record from an email thread Writing	60% Judge: The four-line format, decision and open mezzanine item are correct, but the owner line omits Owen Whitehead's responsibilities to lead the check and send the sheets, and the deadline omits Janet as the recipient of the count sheets.	90% Judge: The four-line record is otherwise accurate, but the Owner line only names Owen Whitehead and does not explicitly state that he leads the stock check and sends the count sheets.

• Decision record from an email thread

  Writing

  Gemini 3.8 Flash

  60%

  Judge: The four-line format, decision and open mezzanine item are correct, but the owner line omits Owen Whitehead's responsibilities to lead the check and send the sheets, and the deadline omits Janet as the recipient of the count sheets.

  GPT-6.1 Sol

  90%

  Judge: The four-line record is otherwise accurate, but the Owner line only names Owen Whitehead and does not explicitly state that he leads the stock check and sends the count sheets.

On everyday prompts

Our first round, on 7 November 2026: 36 everyday prompts asked twice, with the same models and settings.

Scores on 36 everyday prompts, 7 November 2026

Model	Score	Full marks	Typical reply time	Cost of all answers
GPT-6 Astra OpenAI	99.2%	69 of 72	2.6 s	$0.370
GPT-6.1 Sol OpenAI	98.9%	68 of 72	2.6 s	$0.076
Gemini 3.8 Flash Google	98.7%	67 of 72	5.7 s	$0.164
Gemini 4 Argon Google	98.6%	67 of 72	3.9 s	$0.120
GPT-6 Luna OpenAI	98.5%	67 of 72	2.3 s	$0.005

• GPT-6 Astra OpenAI 99.2%

  Full marks

  69 of 72

  Reply time

  2.6 s

  Cost

  $0.370

• GPT-6.1 Sol OpenAI 98.9%

  Full marks

  68 of 72

  Reply time

  2.6 s

  Cost

  $0.076

• Gemini 3.8 Flash Google 98.7%

  Full marks

  67 of 72

  Reply time

  5.7 s

  Cost

  $0.164

• Gemini 4 Argon Google 98.6%

  Full marks

  67 of 72

  Reply time

  3.9 s

  Cost

  $0.120

• GPT-6 Luna OpenAI 98.5%

  Full marks

  67 of 72

  Reply time

  2.3 s

  Cost

  $0.005

All five scored between 98.5% and 99.2%, and every mark they lost was on writing and summaries. The GPT models replied faster.

The 36 harder prompts

Six kinds of task. Checked answers were marked automatically, coding answers were run against hidden tests, and the judge marked the writing against a written scheme.

• Every answer from the harder round
• Every answer from the everyday round
• How we tested

Gemini and GPT models on whichai.chat

The models whichai.chat recommends from each company today, with figures from the live model catalogue.

Recommended Google and OpenAI models on whichai.chat

Model	Company	Context window	Reads images
Gemini 3.8 Flash	Google	1M tokens	Yes
Gemini 4 Argon	Google	1M tokens	Yes
GPT-5.6 Luna	OpenAI	1.1M tokens	Yes
GPT-6 Astra	OpenAI	1.1M tokens	Yes
GPT-6.1 Sol	OpenAI	1.1M tokens	Yes

The list follows the recommended models in the app and changes when new ones are released.

Run the comparison yourself

1. 1

   Add a Gemini and a GPT model

   Start a chat and pick one of each from the model picker. Add Claude or others too if you like, up to ten models in a chat.

2. 2

   Ask your own question

   Use something from your work, with PDFs, text files or images attached if it needs them. Tick Web Search when the answer depends on recent information.

3. 3

   Compare the answers

   Each model's answer has its own tab. Tick Consensus Summary for one model to read them all and write a single comparison.

4. 4

   Keep the one you trust

   Choose which answers your next message builds on and carry on from there.

Questions