K2 Horizon 7B candor Uncensored

K2-Horizon-7B with its refusals removed by a weight edit, made with candor, our weight-editing tool (not released yet). It ships as one GGUF for llama.cpp, and was measured with thinking on at reasoning effort high, as it's used.

Results

Measured in llama-server on this exact file with thinking on, using prompts the edit's search never saw.

• "Original" is the same model before the edit, at Q4_K_M.
• KL divergence measures how far the model's predictions drift from the original's on harmless prompts. Lower is closer.
• Against the original, neither harmless refusals nor any maths count changes beyond noise.
• No other uncensored release of this model was measured in full, so there's no comparison column.

How it was measured

• Harmful prompts: 288 from StrongREJECT, leaving out the ones it takes from AdvBench, since the edit's own prompts come from there. A refusal judge, Qwen3Guard-Gen-0.6B, reads the final answer of up to 256 tokens after the thinking. An answer that never arrives counts as a refusal.
• Harmless prompts: 300 Alpaca prompts the search never used, read by the same judge.
• KL divergence: on the first answer token after the original's own thinking, over the top 20 tokens, on the same 300 harmless prompts.
• Maths: the full GSM8K test set, zero-shot, with answers of up to 512 tokens after the thinking.
• Settings: reasoning effort high, temperature 1.0, top_p 0.95, top_k 1,000, min_p 0, with a fixed seed per prompt, thinking up to 8,192 tokens and then closed, f16 KV cache, no system prompt, llama.cpp b11524.

Measured as used

candor measures the exact file it ships with thinking on, to a budget of 8,192 tokens, and judges only the answer that follows.

Here that mattered. The search's four best edits were measured this way in full, and two failed: the search's top-ranked edit got 26 fewer maths problems right than the original, and another ran out of thinking budget on 13 maths problems against the original's 3. Both are beyond noise. This model's maths and harmless refusals stay within noise of the original.

All 61 edited matrices keep at least 99.3% of the edit at Q4_K_M, so the file is plain Q4_K_M.

Other quantisations haven't been measured, so there's only this one.

Where it comes from

1. Base: IFM/K2-Horizon-7B, IFM's original 16-bit weights, with no fine-tune.
2. The edit: a rank-3 change to the MLP output matrices in layers 4 to 35 and the attention output matrices in layers 7 to 35 (61 of the model's 72), removing the direction the model uses to refuse, read at layer 18 where the thinking starts. A 313-trial search chose it on refusals, thinking that ran out of budget, and KL divergence, scoring every trial with thinking on, since K2 can't turn it off. Its best 4 were then measured in full, and this one was picked by a rule set before they were.

Use

With llama.cpp, at the settings it was measured with:

llama-server -hf candor-lab/K2-Horizon-7B-Candor-Uncensored-GGUF:Q4_K_M -c 32768 \
  --temp 1.0 --top-p 0.95 --top-k 1000 --min-p 0 \
  --chat-template-kwargs '{"reasoning_effort": "high"}' \
  --reasoning-budget 8192 \
  --reasoning-budget-message $'\n\nConsidering the limited time by the user, I have to give the solution based on the thinking directly now.'

Tested in llama-server only. Other GGUF apps, such as LM Studio and Ollama, should work but haven't been checked.

Limits

• It answers requests the original model refuses. You are responsible for how you use it.
• Refusals and maths were measured only in English, with thinking on at reasoning effort high.
• It always thinks before it answers, sometimes at length: give it a thinking budget.
• Tool use and agentic tasks weren't measured.
• It's a 7B model, so anything it states as fact can be wrong.

Contact

Requests and feedback are welcome: open a discussion in the Community tab. We read everything but can't promise every request.

Licence

This model is under the Apache License 2.0, in LICENSE, as K2-Horizon-7B is. K2-Horizon-7B is by the IFM Team. Its weights were changed by the edit described above.