Published by TokForge for on-device inference.

TokForge

• Website: https://tokforge.ai
• Discord: https://discord.gg/Acv3CBtfVm
• Google Play: https://play.google.com/store/apps/details?id=dev.tokforge
• iOS TestFlight: https://testflight.apple.com/join/jnufjzRr

Runs on-device in the TokForge app.

Qwen3.5-35B-A3B-heretic-v2-GGUF

Weighted/imatrix Q2_K GGUF of the heretic v2 abliterated Qwen3.5-35B-A3B MoE, republished under a shorter neutral file name for the TokForge model catalog.

All credit for the model, the abliteration pass, and the quantization goes to the upstream authors listed below. This repo mirrors a single upstream file, renamed only; the bytes are identical.

What this is

• Architecture: Qwen3.5-35B-A3B, a mixture-of-experts model with 35B total parameters and roughly 3B active per token (qwen35moe in llama.cpp terms).
• Abliteration: heretic v2 pass by llmfan46, with the model's native multi-token-prediction (MTP) head preserved through the process.
• Quantization: weighted/imatrix Q2_K (2-bit) from mradermacher's i1 series.
• MTP head: inert for standard autoregressive decoding in current llama.cpp builds, but kept intact in the weights so future speculative decoding support can use it.

Files

sha256: cb60a6e0acbb93c9ab5e85aade11d9ab1fb81ea7e63dcf80123f4656b079a49f

Credits

• Qwen3.5-35B-A3B by the Qwen team (Alibaba Cloud) (Hugging Face). Used under the Apache-2.0 license.
• Qwen3.5-35B-A3B-uncensored-heretic-v2-Native-MTP-Preserved by llmfan46 (Hugging Face). Used under the Apache-2.0 license.
• The GGUF file is byte-identical to the i1 Q2_K quant by mradermacher (Hugging Face), Apache-2.0.
• GGUF format and quantization from llama.cpp (MIT).

License and safety

• Apache-2.0, inherited from the Qwen base model and the llmfan46 abliteration.
• This is a safety-reduced variant: it refuses far less than the base model. Deploy with appropriate user-facing controls and local policy.

Community

• Website: tokforge.ai
• Discord: Join our Discord