What is the difference between Gemini 3 Pro Preview and GPT-5 Codex?

Gemini 3 Pro Preview is developed by Google AI while GPT-5 Codex is developed by OpenAI. You can compare their actual outputs across 48 challenges on RIVAL to see how they differ in practice.

Which is better, Gemini 3 Pro Preview or GPT-5 Codex?

It depends on your use case. Gemini 3 Pro Preview and GPT-5 Codex each have strengths in different areas. RIVAL lets you compare their real outputs side-by-side across 48 challenges so you can judge which fits your needs best.

How can I compare Gemini 3 Pro Preview and GPT-5 Codex on RIVAL?

This page shows a side-by-side comparison of Gemini 3 Pro Preview and GPT-5 Codex across shared challenges. You can vote on which model produced the better output, and Pro users can create custom challenges to test both models with their own prompts.

Rival

Gemini 3 Pro PreviewvsGPT-5 Codex

43 challenges

Why Gemini 3 Pro Preview?

Leads 1 of 1 benchmarks

Stronger on SWE-bench Verified76.2% vs 74.5%

2 months newerNov 2025

Why GPT-5 Codex?

No standout advantages

Favorites

Gemini 3 Pro PreviewGPT-5 Codex

Movie

The Matrix

19998.2

Blade Runner

19827.9

Album

The Dark Side

Gregorian

Kind of Blue

Miles Davis

Book

The Hitch Hiker's Guide to the Galaxy

Douglas Adams

—

City

Kyoto

Japan

Tokyo

Japan

Game

Tetris (1984)

Puzzle4.4

Minecraft

Action, Arcade4.4

12 / 43

Estimate Complexityreasoning

Tests an AI's ability to make educated estimates based on technical knowledge

Gemini 3 Pro Preview

Google · Nov 2025text

No content available for this output.

GPT-5 Codex

Openai · Sep 2025text

Step 1 – Note the basic training quantities

GPT‑3’s largest model has about (P \approx 1.75 \times 10^{11}) trainable parameters (175 billion).
Public descriptions indicate it was trained on roughly (T \approx 3.0 \times 10^{11}) tokens (≈300 billion).

Step 2 – Relate parameters, tokens, and FLOPs

For dense transformers, almost all computation comes from matrix multiplies.
A practical rule (derived by counting multiply–adds in the attention and MLP blocks, and accounting for the backward pass) is:

[ \text{Training FLOPs} \approx 6 \times (\text{number of parameters}) \times (\text{number of tokens}) ]

Here’s the intuition:

Forward pass: Each parameter contributes one multiply and one add → about (2) FLOPs per parameter per token.
Backward pass: Computing gradients with respect to activations and weights requires roughly twice the forward cost (two more matrix multiplies of comparable size).
So total ≈ (3) times the forward cost → (3 \times 2 = 6) FLOPs per parameter per token.

(The softmax, layernorm, activation costs are much smaller and get absorbed into this constant.)

Step 3 – Plug in the GPT‑3 values

[ \text{FLOPs} \approx 6 \times (1.75\times 10^{11}) \times (3.0 \times 10^{11}) ]

[ = 6 \times 5.25 \times 10^{22} = 31.5 \times 10^{22} \approx 3.15 \times 10^{23} \text{ FLOPs} ]

Step 4 – Interpret

So, training GPT‑3 to completion requires on the order of (3 \times 10^{23}) floating-point operations.
That estimate aligns with detailed breakdowns of transformers and is accurate to within tens of percent, which is sufficient for order-of-magnitude comparisons.

Vote

2 / 43

AI Board Game Logicreasoning

Tests an AI's ability to understand game rules and strategy

Gemini 3 Pro Preview

Google · Nov 2025svg

No content available for this output.

Openai · Sep 2025text

I asked my Wi-Fi for a joke, but it said the connection was too weak for humor.
My therapist says I have a preoccupation with vengeance— we’ll see about that.
I tried to start a hot air balloon business, but it never really took off.
I told my coffee it was grounded, now it won’t espresso itself.
I asked the librarian if the library had books on paranoia—she whispered, “They’re right behind you.”

No content available for this output.

Vote

11 / 43

Minimalist Landing Pageweb design

Tests an AI's ability to generate a complete, working landing page

Gemini 3 Pro Preview

Google · Nov 2025website

GPT-5 Codex

Openai · Sep 2025website

Vote

12 / 43

Pokémon Battle UI Recreationweb design

Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.

Gemini 3 Pro Preview

Google · Nov 2025website

No content available for this output.

GPT-5 Codex

Openai · Sep 2025website

No content available for this output.

Vote

12 of 43

FAQ

Continue exploring

Gemini 3 Pro Preview vs Grok 3New provider

Gemini 3 Pro Preview vs Llama 4 MaverickNew provider

Gemini 3 Pro Preview vs Claude Opus 4New provider

Step 1 – Note the basic training quantities

GPT‑3’s largest model has about (P \approx 1.75 \times 10^{11}) trainable parameters (175 billion).
Public descriptions indicate it was trained on roughly (T \approx 3.0 \times 10^{11}) tokens (≈300 billion).

Step 2 – Relate parameters, tokens, and FLOPs

[ \text{Training FLOPs} \approx 6 \times (\text{number of parameters}) \times (\text{number of tokens}) ]

Here’s the intuition:

Forward pass: Each parameter contributes one multiply and one add → about (2) FLOPs per parameter per token.
Backward pass: Computing gradients with respect to activations and weights requires roughly twice the forward cost (two more matrix multiplies of comparable size).
So total ≈ (3) times the forward cost → (3 \times 2 = 6) FLOPs per parameter per token.

(The softmax, layernorm, activation costs are much smaller and get absorbed into this constant.)

Step 3 – Plug in the GPT‑3 values

[ \text{FLOPs} \approx 6 \times (1.75\times 10^{11}) \times (3.0 \times 10^{11}) ]

[ = 6 \times 5.25 \times 10^{22} = 31.5 \times 10^{22} \approx 3.15 \times 10^{23} \text{ FLOPs} ]

Step 4 – Interpret