Sonnet 4.5 vs GPT-5: What I Liked (and Didn't) After Testing Them Out
| 10/3/2025
Getting Started
I’ve been closely following the latest AI models lately from their speed and intelligence to how they tackle real-world problems. GPT-5 arrived with high expectations but then Anthropic launched Sonnet 4.5, earning praise for its quickness and sharp performance in certain areas. I decided to start a blog for fun really and this is the first idea, So The verdict? Both are impressive in their own ways and neither is a clear winner across the board, Objectively. However I do prefer one of them.
Quick Overviwe
Sonnet 4.5 stands out for its speed it reliably completes tasks quickly and accurately on most days. In contrast, GPT-5 tends to take its time but it’s clever and excels at big-picture planning.
Pricing Snapshot
Here’s the pricing, in case you were wondering rates below are per million tokens:
| Model | Input | Output |
|---|---|---|
| GPT-5 | $1.25 | $10.00 |
| Sonnet 4.5 | $3.00 | $15.00 |
GPT-5: Where It Shines
GPT-5 is excellent at following instructions. While it can be a bit slower (for context: 7.5s latency vs Sonnet’s 9.3s, though results swapped in actual tests) its steadiness pays off. Its planning capabilities are especially solid, handling complex tasks with ease, scoring 76.6% in math reasoning compared to Sonnet’s 71.1%. It’s a great choice when precision matters.
On the flip side: the slower pace can feel limiting if you’re pressed for time.
Sonnet 4.5: Highlights
Sonnet 4.5 is all about speed, turning around responses swiftly and maintaining a trustworthy 80% accuracy in my runs. Its performance on benchmarks like GPQA (59.4% vs GPT-5’s 53.6%) gives it a slight edge for certain types of questions. The responsiveness is genuinely refreshing.
That said, Sonnet lags behind in UI design tasks, GPT-5 tends to create cleaner, more intuitive layouts. Also Sonnet is still a bit too jumpy at times, occasionally accepting completely idiotic suggestions.
Takeaways
Consider using both: GPT-5 excels at structured, strategic projects while Sonnet 4.5 is great for quick responses. Neither model is flawless so I’d recommend reviewing outputs carefully especially on creative tasks like UI design.
Wrapping Up
Using these models have been class using numerous tools like Opencode and Cursor . Sonnet 4.5 brings the speed that’s easy to appreciate day-to-day but GPT-5’s thoughtful planning means it holds its ground. If you often multitask it’s worth experimenting with both to see which suits your workflow best.