Have you been trying to figure out if GPT-5 is a game-changer or a total dud? Because I have, and the reactions are so polarized it’s giving me whiplash! Some people are calling it the greatest model ever, while others are sticking with Claude. It’s wild.
Thankfully, I just stumbled upon this awesome video where an AI professional breaks down all of the reactions from across the industry, and I had to share the highlights with you.
📈 The Benchmarks Say It’s #1
Right off the bat, the numbers look incredible. The video’s creator highlights independent benchmarks from Artificial Analysis and the LMSys Arena, and in both cases, GPT-5 is sitting at the top of the leaderboard.
One of the coolest things this innovator shared is that GPT-5 has different “reasoning effort” configurations:
- High: For maximum intelligence on complex tasks.
- Medium: A great balance of smarts and speed.
- Low & Minimal: For when you need answers fast and efficiently.
So according to the data, OpenAI has reclaimed the #1 spot. But as this channel points out, benchmarks aren’t the whole story.
🤔 Are We in a “Post-Eval” Era?
This was a fascinating point the YouTuber brought up from a post by Theo.gg. The idea is that we’re past caring about tiny percentage point wins on benchmarks. What matters now is the vibe of the model.
Does it follow instructions well? Does it feel good to code with? The creator shows that many top developers feel GPT-5 is the best model they’ve ever used for actually getting work done, regardless of what the charts say.
👍 The Good: Smarter, Cheaper, and Less Annoying
The fans of GPT-5 are seriously impressed. The person who shared this video collected a ton of positive takes:
- Better for Agents: One expert ran tests showing GPT-5 is far more reliable for computer use tasks than GPT-4o was.
- Great Personality: Many users love that it’s more direct, to the point, and doesn’t hallucinate as much. It’s not a suck-up!
- One-Shot Power: An intern at LMSys Arena apparently created a working Minecraft clone in a single prompt. Insane!
👎 The Bad: Slower and Not Always Better?
Of course, not everyone is sold. The creator also presented the other side of the argument:
- Worse for Browsing? The team at Stage Hand found that GPT-5 was actually slower and less accurate than Claude’s Opus 4.1 on their browsing evals.
- Still Not Perfect for Code: A developer at Meta shared a hilarious story. The expert explained how GPT-5 beautifully refactored his entire codebase… but none of it actually worked!
- Diminishing Returns: The CEO of Replit feels we’re seeing smaller and smaller improvements, suggesting the hype might be outpacing the innovation.
💰 The Biggest Innovation Might Be Price
I think this is a huge deal. The video’s creator shared a pricing chart, and GPT-5 is shockingly cheap. We’re talking $1.25 per million input tokens. That’s more than 5x cheaper than Claude Opus 4.1 and even cheaper than Claude’s budget model, Sonnet. This makes it so much more accessible for developers and businesses.
This launch has been a rollercoaster, with valid points on all sides. It seems like the best model now really depends on your specific use case.
The expert who made the video goes into way more detail on all of this. For the full deep-dive, make sure to watch the original video from the creator!