OpenAI claims GPT-5 model boosts ChatGPT to 'PhD level' - BBC News, August 7, 2025
Hmm, no. Maybe in a future release, but version 5 of ChatGPT isn't throwing off many PhD vibes.
It’s been less than a week since ChatGPT 5 hit the market, and we have been running it through real-world scenarios. Our verdict? Underwhelming. And that’s not just “we expected more” underwhelming—it’s “really, this is the best they’ve got?” underwhelming. ChatGPT continues to be an awesome tool with many incredible capabilities, but its still got some awkward spots and version 5 came with a lot of over-the-top hype.
Here’s the breakdown, no sugar-coating.
1. The model confusion is gone—but was it even a problem?
Yes, it’s nice that you no longer have to scroll through a buffet of oddly named models (4o, 01 mini, 4.1 nano, etc.). But let’s be honest—most users didn’t even know those options existed, let alone when to pick one over another. For everyday business leaders, this improvement feels like fixing a door no one knew was broken.
2. Web search is still embarrassingly bad.
For a tool that’s supposed to be your “knowledge assistant,” ChatGPT still can’t pull reliable, timely info from the web. Claude and Perplexity make it look effortless—accurate, fast, and easy. Meanwhile, ChatGPT’s “real-time search” ...well, isn't real-time or a search. If you’re in sales, marketing, or leadership, this matters: you can’t rely on it for competitor tracking, market shifts, or live campaign performance without cross-checking everything.
3. The AGI dream is still a dream.
With v5, the gap between hype and reality on reasoning skills is clearer than ever. We were promised “light years” of improvement in complex reasoning—what we got still makes avoidable mistakes on basic logic. It’s a reminder: LLMs are fantastic auto-completion engines, not omniscient business strategists. They’re absolutely great for pattern-matching and text generation, but still bad at reading between the lines, interpreting nuance, or thinking like a human. That’s a reality check for anyone fearing Agentic AI is taking their job tomorrow—it’s not. And for those worrying about the various doomsday AI scenarios, you can put a pause on your anxiety. ChatGPT 5 won't be stealing your job or exterminating humanity any time soon.
4. Coding is the star feature—but only if you code.
OpenAI praised itself for the the advancements it made in coding with v5. That’s great—if you write software. But for the majority of SMB and mid-market leaders, faster Python generation isn’t moving the needle on quarterly revenue. This feels like OpenAI’s play to close the gap with Claude for engineering teams, not to make life dramatically easier for non-technical professionals.
5. Still missing the business-first polish.
For all the advancements, v5 still struggles with tasks business leaders actually care about—like generating sales-ready proposals that don’t require heavy editing, surfacing insights from CRM data without manual prompting gymnastics, or giving context-aware recommendations for strategic decisions. These are the features that would make ChatGPT feel indispensable in the boardroom, and they’re still not there.
Final Thought
ChatGPT 5 is a "nothing burger" for the average business leader. Yes, there are quality-of-life improvements, and yes, coding gets a noticeable bump. But the features that would make it a true game-changer for SMB and mid-market leaders—accurate real-time intelligence, human-level reasoning, and frictionless business application—still aren’t here. The hype cycle will roll on, but for now, business leaders should treat ChatGPT as a helpful assistant, not a strategic oracle.
Why can't ChatGPT do real-time web searches like other LLMs?
So much wrongness in this image. Lizzo and Cardi B are mixed up. Sza, Travis Scott Lana Del Rey are shown twice but with different images. And ChatGPT got confused on showing 20 famous musicians with 10 per row.
