4 ms·
I find it baffling that Google offers such impressive models through the API and even the free AI Studio with fine-grained control, yet the models used in the G
by thimabi 1y ago
I find it baffling that Google offers such impressive models through the API and even the free AI Studio with fine-grained control, yet the models used in the Gemini app feel much worse.
Over the past few weeks, I’ve been using Gemini Advanced on my Workspace account. There, the models think for shorter times, provide shorter outputs, and even their context window is far from the advertised 1 million tokens. It makes me think that Google is intentionally limiting the Gemini app.
Perhaps the goal is to steer users toward the API or AI Studio, with the free tier that involves data collection for training purposes.
- Alifatisk 1y agoGoogle lack marketing for ai studio, it has only recently become widely known through word of mouth
- thimabi 1y agoThat does work in Google’s favor. Users who are technical enough to want a better model eventually learn about AI Studio, while the rest are none the wiser.
- _delirium 1y agoThis might have changed after you posted your comment, but it looks like 2.5 Pro and 2.5 Flash are available in the Gemini app now, both web and mobile.
- thimabi 1y agoOh, I didn’t mean to say that these models were unavailable through the app or website. Rather, I’ve realized that using them through the API or AI Studio yields much better results — even in the free tier. You can check that by trying prompts with complex instructions and long inputs/outputs. For instance, ask Gemini to generate notes from a specific source (say, a book or class transcription). Or ask it to translate a long article, full of idiomatic expressions, while maintaining high fidelity to the source. You will see that the very same Gemini models are underutilized on the app or the website, while their performance is stellar on the API or AI Studio.
- mpalmer 1y agoUnderutilized, or over-prompted for the layperson?