Post

Do you know to what extent "flash" is a description of the model / inherent to the model vs just a choice of how to market different model sizes?

"That pelican looks like it’s in Miami for a crypto conference." 😂

Sounds like they just haven't quite got it ready for release yet, they said next month

Straight out of Hotline Miami.

I fear AI models might be somewhat veblen goods, and that we all somewhat conflate price with quality

it is weird seeing hacker news have fun

So are the models getting more expensive to run in terms of compute? or are the AI labs and Google capacity constrained and taking out larger margins?

the price bump makes local mlx on apple silicon pencil out for routine code-completion calls. context > 200k still goes to the api, but the floor of "reach for gemini" feels higher now

Ok, yeah looking at the AA results. Gemini 3.5 Flash high took 73M tokens to get a score of 55, Opus 4.7 high took 12M tokens to get a score of 52 and Opus 4.7 max took 110M tokens to get a score of 57