When OpenAI announced GPT-5, the pitch was familiar with better reasoning, smarter responses, faster output. This was supposed to be the new crown jewel, the model that would widen the gap between OpenAI and the rest of the AI world. On paper, the numbers looked good. Benchmarks showed small but clear gains. The marketing made sure we knew this was the next big thing.
But once the novelty wore off, many users started asking the same question: Why does this feel… worse?
The answer is subtle. GPT-5 is stronger than its predecessor in some ways ( ~5% increase in performance over o3). It scores slightly higher on technical benchmarks, solves a few more problems, and does so more efficiently. Yet the experience of using it doesn’t match the hype. Conversations feel colder, less human. Inconsistencies creep in. Some days, it nails your query; other days, it feels like you’re talking to a stripped-down version of GPT-4o.
And that’s where the problem lies: expectations vs. lived reality.
The Reality of the Upgrade
GPT-5’s raw performance improvement over o3—the previous best reasoning model is around 5% on coding tasks. That’s respectable in technical circles but far from the leap many hoped for. The biggest change isn’t necessarily the intelligence it’s the routing system that decides when you get the “fast” version and when you get the “thinking” one.
In theory, that should balance speed and quality. In practice, it creates inconsistency. You’re never sure which GPT-5 you’re talking to. Combine that with a personality dialed down to “polite assistant” and you lose much of the spark that made GPT-4o beloved.
This is why so many users call it a downgrade in disguise: technically better, experientially worse.
Meanwhile, Perplexity Is Playing a Different Game
While OpenAI fine-tunes its flagship, Perplexity is making moves that directly expand its reach. The biggest? Partnering with Airtel to give all 360 million subscribers a free year of Perplexity Pro. That’s a ₹17,000-per-year product handed out like a SIM card.
This instantly puts Perplexity in the hands of more Indian users than ChatGPT has globally. For a mobile-first country with cheap data and massive linguistic diversity, it’s a perfect proving ground for AI adoption at scale.
And if that wasn’t enough to get attention, Perplexity made headlines with a $34.5 billion( considering as of now in August 2025, the Perplexity valued at ~$18 Billion ) bid to buy Google Chrome. Most analysts see it as a PR stunt, but it worked it kept Perplexity in the news cycle and painted them as bold, unpredictable, and willing to challenge tech giants on their own turf.
The Takeaway
OpenAI’s GPT-5 is a lesson in how incremental technical gains can feel like stagnation if the user experience doesn’t improve in parallel. Perplexity, on the other hand, is showing how aggressive partnerships and audacious PR can shift the momentum of the AI race.
One is quietly refining its engine.
The other is taking over the highway.
