GPT-5
Nothing here. The model returned empty. We stared at it for a while.
GPT-5 vs GPT-5.4
Compare GPT-5 and GPT-5.4, both from OpenAI, context windows of 400K vs 1.1M, tested across 45 shared challenges. Updated March 2026.
FAQ
Loading share button...
GPT-5vsGPT-5.4
45 fights queued
Why GPT-5?
1.5x cheaper overall$1.25/M in · $10.00/M out
Why GPT-5.4?
2.6x more context1.1M
7 months newerMar 2026
Favorites
GPT-5GPT-5.4
12 / 45
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
GPT-5.4
A standard back-of-the-envelope estimate for training FLOPs is:
[ \text{Training FLOPs} \approx 6 \times N \times D ]
where:
- (N) = number of model parameters
- (D) = number of training tokens
- the factor 6 comes from roughly:
- forward pass: (\sim 2ND)
- backward pass: (\sim 4ND)
This is a common transformer scaling approximation.
1) Plug in GPT-3 size
GPT-3 is known to have about:
[ N = 175 \text{ billion} = 1.75 \times 10^{11} ]
A typical quoted training token count is about:
[ D = 300 \text{ billion} = 3 \times 10^{11} ]
2) Multiply
First compute parameter-tokens product:
[ N \times D = (1.75 \times 10^{11})(3 \times 10^{11}) = 5.25 \times 10^{22} ]
Now multiply by 6:
[ 6 \times 5.25 \times 10^{22} = 31.5 \times 10^{22} = 3.15 \times 10^{23} ]
3) Final estimate
[ \boxed{\text{Training GPT-3 required on the order of } 3 \times 10^{23} \text{ FLOPs}} ]
4) Sanity check
This is usually expressed as a few (10^{23}) FLOPs, which matches the well-known rough estimate.
If you want, I can also derive where the (6ND) rule comes from in more detail.
Vote
2 / 45
AI Board Game Logicreasoning
Tests an AI's ability to understand game rules and strategy
Vote
3 / 45
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
GPT-5.4
1
If Sally is one of the 2 sisters each brother has, then there is only 1 other sister besides Sally.
Vote
4 / 45
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
GPT-5.4
- I told my computer I needed a break, and now it won’t stop sending me KitKat ads.
- Why don’t skeletons fight each other? They don’t have the guts.
- I’m reading a book on anti-gravity. It’s impossible to put down.
- I used to play piano by ear, but now I use my hands.
- Why did the scarecrow win an award? Because he was outstanding in his field.
Unique jokes: 5
Vote
5 / 45
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
Vote
6 / 45
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
Vote
7 / 45
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
Vote
8 / 45
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
Vote
Sponsored
9 / 45
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
Vote
10 / 45
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
Vote
11 / 45
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
Vote
12 / 45
Pokémon Battle UI Recreationweb design
Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.
Vote
The full comparison is right here
33+ head-to-head challenges. All of them judged by real people.
Test any model with your own prompts in Prompt Lab
5 free credits to start. No card required.
By continuing, you agree to Rival's Terms of Service and Privacy Policy
