---

base_model: ressl/MiniMax-M3-uncensored

base_model_relation: quantized

license: other

library_name: llama.cpp

pipeline_tag: text-generation

quantized_by: spiritfather

tags:

• gguf
• minimax-m3
• uncensored
• abliterated
• imatrix
• llama.cpp

---

MiniMax-M3-uncensored — GGUF (imatrix quants)

<p align="center">

🧬 Source model: <a href="https://huggingface.co/ressl/MiniMax-M3-uncensored" target="_blank">ressl/MiniMax-M3-uncensored</a> · base: <a href="https://huggingface.co/MiniMaxAI/MiniMax-M3" target="_blank">MiniMaxAI/MiniMax-M3</a>

<br>

🧮 imatrix + recipe: <a href="https://huggingface.co/AesSedai/MiniMax-M3-GGUF" target="_blank">AesSedai/MiniMax-M3-GGUF</a>

<br>

🛠️ Runs on <a href="https://github.com/ggml-org/llama.cpp" target="_blank">llama.cpp</a> (master)

<br>

📊 Scored on <a href="https://caliperbench.com/m/minimax-m3-uncensored/" target="_blank">CaliperBench</a>

<br>

💬 <a href="https://discord.gg/rhE9zdKyE4" target="_blank">Discord</a>

</p>

> [!Note]

> All credit for the model goes to ressl — abliteration (refusal removal) of o_proj and every down_proj across all 60 layers, routers untouched — and MiniMaxAI for the base MiniMax-M3 (427B total / 10B active MoE, 128 experts, 4 active). These are quantizations only.

> [!Important]

> The importance matrix and the per-tensor quant recipe are AesSedai's, taken verbatim from AesSedai/MiniMax-M3-GGUF (bartowski calibration_datav5, 49 chunks). Each size here is tensor-for-tensor type-matched to their stock MiniMax-M3 quant of the same name, so the only difference from their files is the uncensored weights. Thank you.

> [!Tip]

> Benchmarked on CaliperBench — a creative-writing benchmark scoring prose craft, roleplay and willingness rather than general intelligence. This build is read against the stock MiniMax-M3 served from the same AesSedai recipe.

> imatrix (importance matrix) quants spend their bits where the calibration text says the model is sensitive; below Q5 they are usually preferable to a static quant of the same size. Files over 50 GB are split into shards with llama-gguf-split — point llama.cpp at shard 00001 and it loads the whole set. Text-only; unpruned (all 128 experts kept).

Provided quants

PPL, the PPL-ratio column and KLD are measured against this model's own Q8_0 on wiki.test.raw (512-token chunks, llama-perplexity --no-repack), so they read as quantization loss of the uncensored weights, not a comparison to the stock model. 1-(Mean PPL(Q)/PPL(base)) is the PPL excess over that Q8_0 base (lower is better; Q8_0 is the reference at 0). The table fills in as each size lands.

| Quant | Size | Mixture (dense / gate exps / up exps / down exps) | PPL | 1-(Mean PPL(Q)/PPL(base)) | KLD vs Q8_0 |

|:--|:--|:--|:--|:--|:--|

| IQ4_XS | 189.13 GiB (3.81 BPW) | Q6_K / IQ3_S / IQ3_S / IQ4_XS | 4.5772 | +11.85% | 0.1685 |

| Q5_K_M | 295.21 GiB (5.95 BPW) | Q8_0 / Q5_K / Q5_K / Q6_K | 4.1219 | +0.73% | 0.0377 |

| Q4_K_M | 246.11 GiB (4.96 BPW) | Q8_0 / Q4_K / Q4_K / Q5_K | 4.2198 | +3.12% | 0.0680 |

| IQ3_S | 148.04 GiB (2.98 BPW) | Q6_K / IQ2_S / IQ2_S / IQ3_S | 5.2967 | +29.44% | 0.3703 |

| IQ2_S | 134.01 GiB (2.70 BPW) | Q6_K / IQ2_XS / IQ2_XS / IQ3_XXS | 5.7374 | +40.20% | 0.4698 |

| Q8_0 | 422.46 GiB (8.50 BPW) | Q8_0 (pure) | reference | 0.0000% | reference |

Loading (any current llama.cpp)

llama-server -m IQ4_XS/MiniMax-M3-uncensored-IQ4_XS-00001-of-00005.gguf -ngl 99 -fa on -c 32768 -ctk q8_0 -ctv q8_0 --jinja

The IQ4_XS fits a 256 GB Apple Silicon box.

Thinking: on by default in MiniMax-M3's template. For the non-thinking mode pass chat_template_kwargs: {"thinking_mode": "disabled"} per request.

Quant recipe

1. ressl/MiniMax-M3-uncensored bf16 safetensors → Q8_0 GGUF with convert_hf_to_gguf.py (llama.cpp b10819).
2. Each smaller size cut from that Q8_0 with llama-quantize --allow-requantize using AesSedai's imatrix.gguf and their mixture: the ftype sets the dense tensors and --tensor-type overrides push only the routed experts down. Example for IQ4_XS:

```

llama-quantize --allow-requantize --imatrix imatrix.gguf \

--tensor-type ffn_gate_exps=iq3_s --tensor-type ffn_up_exps=iq3_s --tensor-type ffn_down_exps=iq4_xs \

MiniMax-M3-uncensored-Q8_0.gguf MiniMax-M3-uncensored-IQ4_XS.gguf Q6_K

```

1. Split into <50 GB shards with llama-gguf-split.

License

Inherited from the base model MiniMaxAI/MiniMax-M3.