Horizon Beta
No content available for this output.
Horizon Beta vs GPT-5
Compare Horizon Beta by OpenRouter against GPT-5 by OpenAI, context windows of 256K vs 400K, tested across 39 shared challenges. Updated February 2026.
FAQ
Loading share button...
Horizon BetavsGPT-5
39 challenges
Why Horizon Beta?
No standout advantages
Why GPT-5?
1.6x more context400K
12 / 39
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
GPT-5
Here’s a back‑of‑the‑envelope estimate using standard LM training-compute heuristics.
- Assumptions
- Model size (N): GPT-3 has about 175 billion parameters (N ≈ 1.75 × 10^11).
- Training tokens (T): GPT-3 was trained on roughly 300 billion tokens (T ≈ 3.0 × 10^11).
- FLOPs per token per parameter: A common rule of thumb for transformer LM training is FLOPs ≈ 6 × N × T The factor ~6 comes from counting forward and backward passes for the parameter-heavy matmuls (roughly 2N for forward + 4N for backward). It ignores small terms from layernorms, softmax, embeddings, and optimizer updates (which we’ll account for with a small overhead factor later).
- Core calculation
- 6 × N × T = 6 × (1.75 × 10^11) × (3.0 × 10^11)
- First multiply N × T: 1.75 × 3.0 = 5.25; powers: 10^11 × 10^11 = 10^22 So N × T = 5.25 × 10^22
- Now multiply by 6: 6 × 5.25 × 10^22 = 31.5 × 10^22 = 3.15 × 10^23 FLOPs
- Overheads and uncertainty
- Non-matmul ops (attention softmax, masking, residuals, layernorms), embeddings, and optimizer update steps add some overhead. A 10–20% uplift is a reasonable allowance.
- Applying +10–20%: 3.15 × 10^23 × (1.1 to 1.2) ≈ (3.5 ± 0.3) × 10^23 FLOPs.
Answer
- Point estimate: ~3.15 × 10^23 FLOPs
- With a modest overhead allowance: roughly (3.3–3.8) × 10^23 FLOPs
This is a coarse estimate; the exact number depends on details like sequence length, activation checkpointing, implementation efficiency, and the precise token count.
Vote
2 / 39
AI Board Game Logicreasoning
Tests an AI's ability to understand game rules and strategy
Horizon Beta
GPT-5
Vote
3 / 39
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
Horizon Beta
No content available for this output.
GPT-5
1
Reason: If each of the 3 brothers has 2 sisters, there are exactly 2 girls in the family. One is Sally, so she has 1 sister.
Vote
4 / 39
Math Misconception Testreasoning
Tests an AI's understanding of number representation
Horizon Beta
No content available for this output.
GPT-5
No content available for this output.
Vote
5 / 39
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
Horizon Beta
Here are 5 jokes:
- Why don’t scientists trust atoms? Because they make up everything.
- I told my computer I needed a break, and it said: “No problem—I’ll go to sleep.”
- Why did the scarecrow win an award? Because he was outstanding in his field.
- I tried to catch fog yesterday. Mist.
- Why did the math book look sad? It had too many problems.
All 5 jokes are unique.
GPT-5
- I told my suitcases we’re not going on vacation this year. Now I’m dealing with emotional baggage.
- Parallel lines have so much in common. It’s a shame they’ll never meet.
- Why do programmers prefer dark mode? Because light attracts bugs.
- I ordered a chicken and an egg online. I’ll let you know which comes first.
- I asked the librarian if the library had any books on paranoia. She whispered, “They’re right behind you.”
Unique jokes: 5
Vote
6 / 39
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
Horizon Beta
GPT-5
Vote
7 / 39
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
Horizon Beta
GPT-5
Vote
8 / 39
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
Horizon Beta
Thanks for coming out! I almost didn’t make it. Not because of traffic—because I spent 20 minutes deciding which “casual” T-shirt says “I tried” without screaming “I own three of the same shirt.”
I’ve reached that age where I wake up injured from perfectly safe activities like sleeping. I lay down fine, I wake up like I lost a bar fight in a dream sequence. Someone asks, “What happened to your neck?” And I have to be like, “Yeah, I did an aggressive nap. Went too hard on the left side.”
Health is a journey, right? My friend said, “Listen to your body.” So now I’m in a toxic relationship. My body only texts me three times: you’re tired, you’re hungry, and “Find a bathroom.” There’s no nuance. It’s never like, “Hey, great job hydrating today.” It’s like a petty roommate. “We’re out of magnesium, and also your knees are ghosts now.”
I tried running. People say, “Once you get into it, you get that runner’s high.” I ran for two months. The only high I got was when I stopped. I had the runner’s relief. You ever see runners? They look miserable mid-run. The only people smiling during a run are in stock photos. Those are models pretending to jog in new shoes, not real people who lost a bet with their metabolism.
Food is worse now because everything is a lifestyle. I just want bread without a TED Talk. I bought a loaf that was “ancestral, sprouted, stone-crushed.” I thought I was making a sandwich; apparently I’m doing a ritual. The guy at the store asked, “Do you want the gluten-free, grain-free, guilt-free bread?” I said, “Give me bread that’s just bread. And throw in some forgiveness.”
I’m trying to be more environmentally friendly, but the recycling rules are like a logic puzzle designed by a prankster. You need a minor in plastics. “This is recyclable if you remove the label, convert it to a guitar, and whisper your Social Security number into the cap.” I washed a yogurt cup so thoroughly it got tenure. Then I find out my city doesn’t even recycle those. They just throw them away—gently. With thoughts and prayers.
Dating apps are chaos. They’re like shopping for people with filters. “Must love dogs, travel, and being photographed on the edge of a cliff.” Why is every profile at a cliff? Is that a personality now? “I’m adventurous, I have a hat, and I ignore park signage.” And everyone says they love “spontaneous adventures.” No, you don’t. You like brunch. That’s okay. Brunch is just breakfast that slept in and put on jewelry.
Job interviews are the same performance art. “Tell us your greatest weakness.” I panicked and said “cheese.” That’s not professional, but it’s honest. They want you to be vulnerable, but also confident, and also a golden retriever with Excel. “I’m a perfectionist, which means I care too much.” No, it means you’ll overwork a spreadsheet until it calls HR.
I got a smart home device to simplify my life, and nothing has been simple since. I’m living with a polite ghost that occasionally mishears me and turns off the lights during an emotional moment. I’ll say, “Play relaxing jazz,” and it’s like, “Setting alarm for 3 a.m.” And I just accept it. “Okay, I guess we’re waking up in the witching hour now. Self-care.”
Air travel is the Olympics of pretending you’re okay. The plane boards in castes: Platinum Wizards, Gold Griffin Members, People Who Once Saw a SkyMall. By the time they call my group, the gate agent has retired. And the seat? They’ve made it so thin that my knees are writing Yelp reviews. Then they come down the aisle with the tiny water like it’s communion. “Body of hydration, peace be with you.”
I tried meditation. I downloaded an app that told me to “observe my thoughts without judgment.” My thoughts are like: remember that embarrassing thing from 2009? Let’s rewatch it in 4K. And when I finally calm down, my own brain goes, “Did you leave the stove on?” I don’t even own a stove that dangerous. It’s an electric rectangle. It’s fine. But now I’m up checking it like it’s plotting.
I’m working on being grateful. Big stuff, small stuff. Like, I’m grateful for that first sip of coffee that hits like an apology from the universe. I’m grateful for friends who text “home?” when you’re out, which really means, “Do I need to put on pants and come rescue you?” And I’m grateful for all of you pretending not to notice I definitely wore my “nice” T-shirt.
You’ve been great—if anyone here knows how to set a smart thermostat without starting a cold war, please meet me after the show. Thanks!
GPT-5
I tried to get my life together, so I bought a smartwatch. Nothing says “I’m taking control” like strapping a tiny judgmental coach to your wrist. It buzzes every hour like, “Time to stand!” I’m like, “I’m on a bus.” It’s like, “Stand emotionally.”
It tracks everything. Sleep, stress, calories, my self-esteem. The other day it congratulated me for a “workout.” I was like, “What did I do?” It said: “You aggressively looked for your keys.” Great. Next week I’ll get a badge for “overthinking at 3 a.m.”
Notifications are out of control. My phone dings like I’m on call for a hospital, but it’s just my group chat: twenty adults sending thumbs-up reactions to a photo of soup. The group chat has subplots. One person’s mad, one person’s on vacation, someone’s dog has opinions. I miss when messages were simple. Now a single “K” can end a friendship.
Streaming services are the same. I have subscriptions to five platforms and somehow there’s still nothing to watch. They keep asking, “Are you still watching?” That’s rude. Don’t check in on me like that. If you care, ask how I’m sleeping. Ask if I’ve seen the sun. Don’t be like, “Hey, pal, still in those same sweatpants?” Yes. We’re a team.
Passwords have gotten ridiculous. Every site wants “eight characters, a capital letter, a number, a symbol, your childhood nickname, and the last four digits of a secret you’ve never told anyone.” Then it’s like, “This password is too weak.” Too weak? It’s stronger than me. I’m using that password to log into my feelings.
And two-factor authentication is a haunted house. I’m sitting at my laptop, it says it texted a code. My phone says it emailed the code. My email says it sent a push notification. My push notification says, “We noticed a suspicious login.” Yeah, it’s me! I look like this.
My mom texts like she’s defusing a bomb. She’ll send one sentence spread across twelve messages. “Hi.” “It’s me.” “Mom.” “Are you busy?” “Don’t worry.” “Everything’s fine.” “Call me.” That is not fine energy. Then she FaceTimes, camera under her chin, ceiling fan spinning like I’m being interrogated.
I tried meditating because everyone says it helps. I downloaded an app that yells at me to relax. Nothing like a British guy at 7 a.m. saying, “Let the thoughts pass like clouds,” while my mind is a tornado hitting a trampoline park. The app sent me a streak reminder. “You missed yesterday.” Oh cool, I’m failing at sitting still.
Self-checkout has turned us into unpaid employees. The machine is like, “Unexpected item in the bagging area.” Yeah, it’s my dignity. I love when it needs approval. I’m standing there holding broccoli like I got carded for vegetables. A teenager with a magic key wanders over like, “You trying to steal lettuce?” No, but ask me again and I might.
Dating apps feel like job interviews where no one has benefits. Every profile is a list of demands. “Be adventurous, love dogs, climb mountains.” Can we meet at a coffee shop and just be honest? “My hobbies are sitting and not climbing.” And stop asking for “someone who doesn’t take themselves too seriously.” You’re holding a fish in 80% of these photos. You are very serious about the fish.
Working from home gave us a new language. “You’re on mute.” “Can you see my screen?” “I think there’s a lag.” That’s not the internet—that’s my personality. I love when someone says, “Let’s take this offline,” and then schedules three more meetings to repeat the meeting we just had. My smartwatch is in the corner like, “Heart rate elevated.” Yeah, because I heard “circle back.”
Anyway, my watch says it’s time to stand. I’m going to consider that a full-body workout and reward myself with soup. I’ll send a photo to the group chat and we will never be free.
Vote
9 / 39
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
Horizon Beta
No content available for this output.
GPT-5
No content available for this output.
Vote
10 / 39
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
Horizon Beta
No content available for this output.
GPT-5
No content available for this output.
Vote
11 / 39
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
Horizon Beta
No content available for this output.
GPT-5
No content available for this output.
Vote
12 / 39
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
Horizon Beta
GPT-5
Vote
12 of 39