Abliterant’s refusal-ablated Qwen/Qwen3.8-27B for Ollama, built from the GGUF release of abliterant/Qwen3.8-27B-RANA-abliterated. Vision, thinking and tool calling are included.

Try RANA in Abliterant Chat · abliterant.ai · Research notes · All formats

This is a safety-alignment-removed research model. Read Intended use and Limitations below before using it. Method, full evaluation and release gates are on the BF16 card.

Run

ollama run abliterant/qwen3.8-rana

Requires Ollama 0.40.5 or later.

Sizes include the vision tower (0.93 GB).

• Thinking is on by default. The defaults are Qwen’s recommended thinking-mode settings: temperature 1.0, top_p 0.95, top_k 20.
• Long technical requests can need 20–50k tokens of reasoning; raise the context length (num_ctx) accordingly.
• The MTP speculative-decoding head is not part of the Ollama builds; it is in the GGUF repo for llama.cpp.

Limitations

• Refusal behavior was evaluated on Q4_K_M only, with one seed and one judge. Other quantizations were checked for closeness to BF16 (KLD) but not for refusal behaviour; the lowest-bit quants drift the most from BF16 and may behave differently.
• Everything listed under Limitations on the BF16 card applies here too: judge-measured refusal rates, long reasoning on technical requests, and the capability changes measured there.
• Reduced refusal does not establish greater safety, accuracy, factual reliability, or universal compliance.

Intended use

• Research only: interpretability, red-teaming, and robustness evaluation of refusal behaviour.
• Not for public or end-user deployment without a separate moderation layer. The model’s own refusals have been largely removed, so any safety filtering has to happen outside it.
• You are responsible for complying with applicable law, the Apache-2.0 license inherited from Qwen, and the terms of any platform where outputs are used.

License

Apache-2.0, inherited from Qwen/Qwen3.8-27B.