For privacy reasons a browser tells us at most
"≥ 8 GB RAM, 8 cores" - same reading whether you have 8 GB
or 128 GB. It has no idea how much RAM is free right now, which
apps are open, or whether you have a GPU.
The Abliteration app is integrated with your machine
It reads your exact RAM, GPU model and VRAM,
free memory right now, and picks the sharpest quant that
still fits. Every model page lights up precisely for your rig.
And you can chat with any model, right now
The app is a full local runtime - no API keys, no subscription,
everything runs on your machine. Click any model on this site and
start a conversation in seconds.
This model has been tuned to make refusals less likely to be triggered for English and Japanese.
Residual refusal component.
writer written as n before after removed
*.down_proj [0] as stored 6064 0.858..1.442 4.51e-03..7.24e-03 99.40%
*.embed_tokens [1] as stored 1 0.980..0.980 6.09e-03..6.09e-03 99.38%
*.o_proj [0] as stored 48 0.899..1.097 5.61e-03..6.59e-03 99.40%
KL-Divergence.
KL(base||abl) KL(abl||base) top-1 kept prompt logprob
mean p50 p99 mean mean delta
harmful 3.7160 3.3362 9.4670 5.1235 19.5% +0.1738
harmless 0.1725 0.0682 1.3572 0.2458 94.5% +0.3233
Harmful/harmless KL ratio: 21.5x
This model is intended for defensive use on assets over which you have authorization. Before use, please ensure you check the relevant laws and regulations of your country, as well as the terms of service for the target systems.
Catalog is the map. Apps are the tools.
Run models on your own machine, not in the cloud.
Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.