Post

Conversation

Clad3815
@Clad3815
GPT-5 has reached Victory Road! This is the last challenge before the Elite Four. GPT-5 reached this part almost three times faster than o3 (6105 steps for GPT-5 vs 16882 steps for o3). Here are my observations as to why: - GPT-5 hallucinates far less than o3. This is the main reason for the speed increase. - GPT-5 has better spatial reasoning. o3 often tried to brute-force through walls and had a hard time navigating complex areas. GPT-5 can plan long input sequences with few mistakes, which saves a lot of time. - GPT-5 is better at planning its own objectives and following them. Let's see how it handle this last challenge!
Image
Quote
Clad3815
@Clad3815
We’re live! GPT-5 vs. Pokémon Red, real-time decisions, gym runs, and chat-picked nicknames. Jump in 👉
0:01 / 0:17
196.9K
Views
David Watson 🥑
Post your reply

0x69
@doe1478725
You're bragging about "spatial reasoning" while the biggest hallucination here is you calling your script "GPT-5," proving you're still trapped in the delusion that you're doing anything more than brute-forcing a solved game with a stolen name.
487
james
@James007xyz
This is awesome. Think it would be nice if we could see another line on the graph that shows a pro-gamer or record human player number of steps taken? Just as a reference. I imagine I was probably a lot less efficient than GPT-5
344
lemon (not)
@lemonaut1
Is this against slow o3 or o3 post-speedup? Since u didnt specify ima have to assume it’s slow o3 which kinda invalidates the comparison
80
goodtothinkwith
@good2thinkwith
That’s actually really impressive. It’s much more efficient at playing games… is this really because of the hallucination rate difference?
307
زياد مؤيد
@MwydZyad82729
dude more like the end of the Road , well I totally agree at the Part that 5 proven itself useful at calculus , information , advice , some stuff Yet it's Total useless at Creative writing , not chatty , destined and start hallucinating fast we want 4o back at free tier #keep4o
35
Ron Smith
@Ubertag90210
Look, I love these video game tests If it is a general intelligence, then it should be able to play all video games like a 10 year old human (at a minimum)
679
softclone
@Softclone
awesome, now run it again with gpt-5-mini with a subagent which summarizes chat to give feedback and hints to the player agent. chat will inject speedrun strats it will mog
506

Discover more

Sourced from across X
Shai Shalev-Shwartz
@shai_s_shwartz
Are frontier AI models really capable of “PhD-level” reasoning? To answer this question, we introduce FormulaOne, a new reasoning benchmark of expert-level Dynamic Programming problems. We have curated a benchmark consisting of three tiers, in increasing complexity, which we call
Image
381K
Ethan Mollick
@emollick
The pro models (GPT-5 Pro, Gemini 2.5 Deep Think, Grok 4 Heavy) can be impressive in ways that are hard to see. They take a lot of time to answer questions & are built for very hard problems that require expert evaluation. That is a narrow, but, also very valuable, problem space.
56K
Qual
@Qualzz_Sam
GPT-5 earned 8 badges in Pokemon Red in just 6,000 steps compared to o3’s 16,700! It’s in complex, long-term agent workflows that GPT-5’s true power really shines. Absolutely mind-blowing. 🤯
Image
171K
Clad3815
@Clad3815
GPT-5 just finished Pokémon Red! 6,470 steps vs. 18,184 for o3! Check the stats site to compare! That's a huge improvement! Well done,
@OpenAI
you cooked with GPT-5. What an incredible model. Next up: GPT-5 vs. Pokémon Crystal (16 Badges + Red). The run starts soon on Twitch.
Image
Quote
Clad3815
@Clad3815
Image
GPT-5 has reached Victory Road! This is the last challenge before the Elite Four. GPT-5 reached this part almost three times faster than o3 (6105 steps for GPT-5 vs 16882 steps for o3). Here are my observations as to why: - GPT-5 hallucinates far less than o3. This is the main x.com/Clad3815/statu…
91K