Gemini 2.5 Pro’s blatant mistakes and confidence in its claims are hilarious, highlighting corporate greed over customer needs.
I thought Google was going to pull ahead after their Gemini 2.5 Pro 03-25 release, which was the best model they released and how even with 1 million context, it barely hallucinated. I still don’t understand why they deprecated it. This feels like a classic case of a large corporation becoming disconnected from its core users and pushing forward without even trying to get feedback.
Competitors such as Anthropic and OpenAI value user feedback. Anthropic was informed that their models create unnecessary mocks for features and reward hacking during testing, which they addressed with their sonnet and opus 4 models. OpenAI prioritizes user satisfaction, reintroducing 4o and addressing sycophancy issues.
Google, on the other hand, continues to go down their own path with no direction, just a bunch of executives making decisions to maximize profits while delivering a subpar product. I will not be using Gemini APIs in the future for any of my businesses, and I will make it a point to share how disappointing my experience has been.