Gemini 2.0 Flash Thinking beats DeepSeek-R1. Here's why.

Показать описание

Gemini 2.0 Flash Thinking beats DeepSeek-R1. Here’s why.

Gemini 2.0 Flash Thinking Experimental 01-21 has a 1M tokens context window, which is a massive upgrade from the previous max window of just 32,000 tokens. DeepSeek-R1, on the other hand, has a token context window of just 128,000.

It’s also worth noting that Gemini 2.0 Flash Thinking Experimental 01-21 is free to use in Google AI Studio, whereas DeepSeek-R1 costs 55 cents per million input tokens and $2.19 per million output tokens.

One of the most impressive things about Gemini 2.0 Flash Thinking Experimental is that it comes with tool use within its chain of thought, e.g. it can run Python scripts in its thinking process.

DeepSeek-R1, on the other hand, is only able to implement tool use in its final output, meaning that its reasoning capabilities are inferior to those of Gemini 2.0 Flash Thinking.

Рекомендации по теме

Комментарии

Everytime there's a ad the next thing Joe says is 100% true

johnfeiss

Gemini Thinking with Search is my favorite ❤ It automatically determines if a search is needed before answering.

guava_

The only good argument is the last one

gustavo.zermeno

I got better results than R1 on coding questions. DeepSeek showed me more variance or annealing, which is not good in this case. Prompting right matters, but Gemini gave me variants without asking... such as iterative stack, recursive w function, heap buffer. The API is well documented, so I stick with it. I don't have to use it locally. The DeepSeek phone client chokes, server busy. I think the DeepThink was annoying after awhile. I was nonplussed by the hype.

SvarturVetur