Here's a pretty shocking comparison between Cline-pass and Openrouter.
Over the past few days, I used Deepseek 4.1 flash to complete a few tasks, in an environment where I could only use OpenRouter. Take a look at my usage from October 2-3 (the highlighted bar on the chart was $3.07, but together with the next bars, the total was $4.92):
https://com-pute.com/nick/deepseek41f-usage-on-openrouter.jpg
In contrast, over the longer period from October 1-6, I used the exact same model via the Cline-pass API, in several environments where I had the Cline-pass account set up, to complete what I'd estimate to be at least 4x the amount of work:
https://com-pute.com/nick/deepseek41f-usge-on-clinepass.jpg
So, it cost me nearly as much to perform a few tasks via OpenRouter, as I spend for the entire month of use on Cline-pass, BUT I only consumed 1% (1/100) of my monthly usage on Cline-pass (which costs less than $7 per month), to perform at least 4x the volume of work. That means, in this case, Cline-pass cost me something in the order of 1/400 what OpenRouter cost - for the same model, used during the same general business hours.
Don't expect those numbers to represent the norm, or to even be reproducible, but what I can say without reservation is that I use Deepseek v4.1 constantly, every day on Cline-pass, to complete very complex workflows. I mean, I use it A LOT - much more than most typical users ever would - and I never hit any of their rate limits, daily, weekly or monthly. That's been going on for many months, including after Deepseek raised their rates many times higher than they had been previously.
I should also point out that I've been watching Cline-pass typically run Deepseek v4.1 Flash at 200-250 tokens per second, and sometimes over 300 tokens per second 😲
I can also tell you that I use OpenRouter very sparingly now, mostly just to test new models when they come out. That typically involves some short chat interactions, but hardly anything even close to a small project. In contrast, on Cline-pass I'm constantly banging out endless significant projects - writing code constantly, compiling repositories from source, installing software, configuring OS settings, researching and implementing framework solutions to cluster and run new LLMs, etc. With the tiny bit of experimental use on OpenRouter, I almost always spend more money every week, than I do for my entire volume of monthly Cline-pass usage, and I still typically only use a small fraction of my allotted Cline-pass usage. It feels endless.
That's why Cline-pass has been my favorite provider. It performs great, gives me a great selection of recent models, and basically seems free. In reality, I pay $80 once a year (equivalent to 4 months of ChatGPT) - that's all, for a truly massive volume of world class intelligence, via an API that can be hooked up to virtually any agent :)
It certainly feels like I'm cheating the system somehow.
