Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · darkc0de

Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-heretic

Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-heretic is a 27.36B-parameter Qwen image-text-to-text derivative from darkc0de. According to the model card, it is a Heretic-decensored reasoning distill.

Publisher
darkc0de
Task
image-text-to-text
Model type
qwen3_5
License
apache-2.0
Library
transformers
Publication status
Approved for indexing

Model overview

Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-heretic is published by darkc0de as an image-text-to-text model. The captured configuration identifies Qwen3_5ForConditionalGeneration with model type qwen3_5 and Safetensors metadata reports 27,356,728,560 parameters. According to the model card, it decensors TeichAI/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2.

Recorded capabilities

Heretic decensoring

According to the model card, the release is decensored with Heretic v1.3.0 and described as reproducible, with a reproduce directory README.

Reported refusal and KL figures

According to the model card, the publisher reports 8/100 refusals against 98/100 for the original and 0.0774 KL divergence.

Reasoning-distill lineage

According to the model card and tags, the upstream reasoning-distill training cites TeichAI Claude-Opus reasoning datasets and a Qwen3.6-27B base.

Use cases in the source record

  • Image-text-to-text and general text-generation experiments of the kind the publisher lists, including coding and creative writing.
  • Decensored-model research comparing refusal behavior and publisher-reported benchmark rows, using the card's abliteration parameters.

Limitations and unknowns

  • According to the model card, benchmark and refusal figures are publisher- or third-party-reported; no independent Ethen evaluation was extracted.
  • No context-window value was extracted from this record.
  • No verified training-hardware or full training-procedure detail was extracted beyond the card's dataset and abliteration notes.
  • Provider state is historical snapshot data, not independently refreshed current availability.

Source and provenance

Source: darkc0de/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-heretic

Captured: Unknown. Processed: 2026-09-07T19:35:55.229529+00:00.

This is a decensored version of TeichAI/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2 , made using Heretic v1.3.0 This model is reproducible! See the README in the reproduce directory for more information. Abliteration parameters Parameter Value direction_index 41.77 attn.o_proj.max_weight 1.22 attn.o_proj.max_weight_position 51.32 attn.o_proj.min_weight 1.21 attn.o_proj.min_weight_distance 32.79 mlp.down_proj.max_weight 1.44 mlp.down_proj.max_weight_position 42.02 mlp.down_proj.min_weight 0.13 mlp.down_proj.min_weight_distance 37.15 Performance Metric This model Original model ( TeichAI/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2 )…

F001F002F003F004F005F006F007F009F010F011F012F013F014