SAIFIINDUSTRIES/Abliterated-Dolphin3.0-R1-Mistral-24B

TEXT GENERATIONPricing:Input $0.7 / Cached $0.04 / Output $1.16Concurrent Unit Cost:2Model Size:24BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 3, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

SAIFIINDUSTRIES/Abliterated-Dolphin3.0-R1-Mistral-24B is a 24 billion parameter language model based on the Dolphin3.0-R1-Mistral-24B architecture, modified using an "abliteration" technique. This experimental model was created to explore the effects of removing censorship and redirecting refusals, resulting in a significantly "dumbed down" version compared to its base. It is not intended for meaningful use and carries high risks due to its altered state.

Loading preview...

Abliterated-Dolphin3.0-R1-Mistral-24B Overview

This model, developed by SAIFIINDUSTRIES, is an experimental 24 billion parameter language model derived from the Dolphin3.0-R1-Mistral-24B base. It was created through a process inspired by "abliteration" techniques, aiming to investigate the removal of censorship and redirection of model refusals. The primary purpose was curiosity, exploring the effects of such modifications on an already relatively permissive model.

Key Characteristics

  • Experimental Nature: This model is a result of an exploratory process, not designed for practical application.
  • Modified from Dolphin3.0-R1-Mistral-24B: It uses the Dolphin architecture as its foundation but has undergone significant alterations.
  • "Dumbed Down" Performance: Due to the abliteration process and the absence of subsequent DPO fine-tuning to repair potential damage, the model's capabilities are substantially reduced compared to the original Dolphin model.

Good For

  • Research into Abliteration: Useful for those studying the effects of abliteration techniques on language models and their impact on model behavior and performance.
  • Understanding Model Refusals: Provides an example of a model where refusal mechanisms have been intentionally altered.

Limitations and Risks

  • Not for Meaningful Use: The creator explicitly states that this model is not suitable for any practical or meaningful applications.
  • High Risks: Users should be aware of significant risks associated with its use, and the original Dolphin3.0-R1-Mistral-24B is recommended for general purposes.
  • Reduced Intelligence: It exhibits significantly lower intelligence and capability compared to its base model.

Finetunes
1 models