The whole concept behind Anubis is flawed. It tries to block access by imposing a compute cost to block people that are using incredible amounts of compute just to generate and parse the requests.
The people that Anubis tries to block have tons of compute to spare.
For example at runpod, if you rent a container with just one B300, you get 32 cpu cores and 250 gbs or ram that are essentially just sitting there while the gpu does all the work. If you think you can impose a compute cost on these people without blocking everyone else, well think again.
Or in my case, I am running qwen3.8 at home on a couple of gpus, these are attached on 32 core epyc server with 128gb of ram I am pretty sure i have more compute than the typical dev laptop.
I am of course nice, and don't aggressively scrape peoples services.
The people that Anubis tries to block have tons of compute to spare.
For example at runpod, if you rent a container with just one B300, you get 32 cpu cores and 250 gbs or ram that are essentially just sitting there while the gpu does all the work. If you think you can impose a compute cost on these people without blocking everyone else, well think again.
Or in my case, I am running qwen3.8 at home on a couple of gpus, these are attached on 32 core epyc server with 128gb of ram I am pretty sure i have more compute than the typical dev laptop.
I am of course nice, and don't aggressively scrape peoples services.