Gemini CLI performance greatly degraded

[Luisfrdzgr]

As stated in the title, the performance of the CLI has degraded significantly across the board. Back in version 0.1, just a couple of months ago, the system was excellent. It could review large codebases and generate detailed reports covering many aspects of the code. Although it took time to produce responses, they were consistently accurate and insightful.

In contrast, the current version performs poorly. Most of the time, it suggests incorrect solutions to problems or provides brief, insufficient answers. While this decline could be related to a model change on Google’s side, I’ve also noticed issues such as poor formatting, incomplete bullet lists, and even entirely empty responses—issues I have never encountered with the web version.

For example, look at this ouput. Completely empty:

Or in this case, really short answer to a pretty open-ended question, and with missing points (2,3,5). Also notice the ### markdown title indicator not being properly rendered:

Has anybody else experienced similar degradation? It's gone from being the best free CLI (when in the quota), to almost unusable. Do people using API key and paying for the tokens experience better usage?

[fann-z]

我也遇到同样的问题，以前的回答非常的深入完整，并且格式美观，现在的回答超级简洁、格式混乱

[Luisfrdzgr]

Just a new example:

[fann-z]

清空所有缓存，重新安装0.1.22版本就好了，我的已经可以正常使用了

[keyboqrd]

it's the problem of rendering markdowns, just ask gemini to respond with plain text instead.