Hacker News Reader

Gemini 2.0 Experimental is now a leading LLM. They started out poorly, but after a more reasonable 1.5 Pro, 2.0 is in another class entirely and a direct competitor to o1 (or o1-mini as for Gemini 2.0 Flash). They've made quick strides forward as the DeepMind team is kicking into gear, and I feel like they're neglected a bit too often these days, especially now while usage cost on AI Studio is a nice $0.

Definitely. The 1206 is probably my favorite out of any I've ever used.

The Gemini launch was a complete disaster, but technically speaking since February 2024, Gemini 1.5 pro and the ensuing lineup have been very impressive.

A lot can change in a year.

Gemini 1.5 pro is extremely impressive at reading 2 million tokens of a document and answering questions about it. And at least for the time being it's offered for free on AI studio.

imo gemini-exp-1206 is the best public LLM that exists right now.

Gemini-exp-1206 seems significantly dumber than 1.5 pro on reading and comprehending large context documents.