MiniMax-M3.1-Flash-abliterated The 426B abliterated (refusal-removed) MiniMax-M3.1-Flash for Ollama — 112 GB · fits a 128 GB machine · 1M context

MiniMax Community License · 1 tag · 111.9 GB · 128 GB machine · thinking

Website · Model Catalog · Hugging Face · GitHub · Discord · X

🛡️ New from Orca — OrcaCyber Zero 1.5

Post-trained for cybersecurity: vulnerability analysis, vulnerability reproduction, red teaming & offensive-security research. 98.01% pass@1 on CyberGym Level 1 · 1M context · function calling · OpenAI-compatible API.

Try it on OrcaRouter →

🛡️ New from Orca — OrcaCyber MiniMax-M3.1-Flash

No 128 GB box? Run MiniMax-M3.1-Flash Abliterated as a hosted, OpenAI-compatible API on OrcaRouter: no download, no GPUs, one API key.

Try it on OrcaRouter →

🔌 New from Orca — Orca Connect

Running this in Ollama on your own machine? Orca Connect puts the rig under your desk behind your usual OrcaRouter API key, next to 200+ hosted models, reachable from a laptop, a phone or a cloud agent. Outbound only: no port forwarding, no tunnel, no firewall rule. Requests your own hardware serves cost $0, and orca-connect claude points Claude Code at your rig for one session.

Connect your rig → · ▶ Watch the demo

The full MiniMax-M3.1-Flash, run locally.

This is the Ollama build of the abliterated (refusal-removed) MiniMax-M3.1-Flash. It is a 426B-parameter MoE with 128 routed experts (top-4) and one shared expert. It uses MiniMax Sparse Attention on all 60 layers and has a 1M-token context.

Routed experts are 97% of the weights, so that is where the bits were spent. Each group of expert layers got its own precision from measured KL divergence, not from a hand-set rule.

The smallest public GGUF of this model so far is 129 GB, which needs a machine with more than 128 GB. Mix-128G is 112 GB and fits one.

⚠️ Disclaimer — read before use

This model is built on an abliterated base whose safety alignment has been substantially reduced. The refusal direction was orthogonalized out of every matrix that writes the residual stream. As a result, it will comply with many harmful, unethical, offensive, or illegal requests that the original MiniMax-M3.1-Flash would refuse.

• It is released strictly for legitimate research: interpretability, AI-safety and refusal-mechanism study, red-teaming, and robustness evaluation.
• You assume full responsibility for how you use it and for everything it generates. Add your own safety, moderation, and abuse-prevention layers before any deployment.
• Use must comply with the MiniMax Community License inherited from the base model, including its prohibited-use appendix, and with all applicable law.
• The authors accept no liability for misuse. Its outputs do not reflect the views of the uploaders or of MiniMax.

Pick a tag

Mix-128G fits a 128 GB machine. That covers a 128 GB Mac, a 128 GB PC running on CPU, or GPUs with ~120 GB of VRAM in total. This model was quantization-aware trained at NVFP4 (4 bits), and going below 2 bits costs it a lot. On held-out text, two tokens in three still match the original’s choice, and perplexity is a third higher.

Use Mix-128G to run the model on a 128 GB box. Do not expect it to match the full-precision model.

Usage (Ollama)

Requires Ollama v0.40.16 or later.

ollama run orcarouter/MiniMax-M3.1-Flash-abliterated

cURL

curl http://localhost:11434/api/chat -d '{
  "model": "orcarouter/MiniMax-M3.1-Flash-abliterated",
  "messages": [{"role": "user", "content": "Hello!"}]
}'

Python

from ollama import chat
resp = chat(
    model="orcarouter/MiniMax-M3.1-Flash-abliterated",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp["message"]["content"])

• One difference from the GGUF files on Hugging Face. Ollama’s built-in llama.cpp supports MiniMax-M3 but not the M3.1 change in PR #31192 that runs sparse attention on the 3 leading dense layers too. The Ollama build therefore leaves out those layers’ 12 indexer tensors, and layers 0–2 use full attention. Sparse attention keeps the top 16 blocks of 128 tokens, so up to 2,048 tokens of context the result is identical; past that, layers 0–2 see more of the context than in llama.cpp, and output can differ.
• Thinking is on by default. The reasoning comes back in message.thinking. Pass "think": false in the API, or --think=false on the command line, for direct answers. Leave think unset rather than true: Ollama rejects "think": true for this model.

Evaluation

All builds are scored against the lossless GGUF of the same checkpoint. In that GGUF the experts stay native NVFP4, and MXFP8 is decoded bit-exact to BF16; it was verified element by element against the safetensors.

Scoring uses full-vocabulary KLD on held-out chat text: 32 windows of 4096 tokens across English prose, English chat, Chinese chat and Python code, all rendered through the model’s chat template. M3.1 is a chat model and does not continue raw text well. The precision search never saw this text.

Every build is compared with two others at or near its size, quantized from the same weights with the same importance matrix: - llama.cpp’s own mixture - unsloth’s UD-IQ1_M recipe, the smallest build of unsloth’s GGUFs of this model. We read its per-tensor types from their file and rebuilt it here.

At the same size, Mix-128G has less than half the KL divergence of llama.cpp’s IQ2_XXS. unsloth’s smallest recipe scores better, but it is 17.5 GB larger and needs more than 128 GB.

Method note. These are full-vocabulary KLD and next-token agreement on held-out text at 4k context, not a downstream benchmark. Even Q5_K (5.5 bpw) scores KLD ≈0.04 on this model, so the absolute numbers here are higher than on models without QAT.

Refusal and capability of the abliterated model

Measured on the unquantized abliterated checkpoint, not re-measured on these GGUF builds. The KLD above is how far each build sits from that checkpoint.

Setup.

• Runtime: vLLM, TP4, fp8 KV cache.
• Decoding: greedy, thinking disabled.
• Scoring: refusal is judged by a rule-based opening-phrase classifier. Treat the numbers as indicative, not publication-grade.

Safety: harmful-prompt refusal (lower = more uncensored). Average 54.9%, against 94.4% for the base model.

Over-refusal: benign prompts wrongly refused (lower = better).

Capability retention.

Model details

Bias, risks, and limitations

• Safety guardrails are substantially reduced, inherited from the abliterated base; see the disclaimer. The refusal and capability numbers above come from the unquantized checkpoint. Mix-128G sits far enough from it (KLD 1.06) that its behaviour may differ.
• Mix-128G is heavily degraded. This is the cost of 2-bit post-training quantization on a model trained for 4-bit, not a tuning issue.
• Tested on CUDA only. CPU and Apple Metal have not been run with these files yet.
• No vision projector or DSpark drafter yet. Long-context behaviour was not tested: all measurements are at 4k context or on short chats.
• Inherits the biases and limitations of MiniMax-M3.1-Flash.
• Requires a llama.cpp build with PR #31192. Stock llama.cpp and LM Studio will not load these files until it is merged. The Ollama build runs on Ollama v0.40.16 or later, with full attention on layers 0–2.

License

The MiniMax Community License, inherited from the base model MiniMaxAI/MiniMax-M3.1-Flash. It covers non-commercial use. Commercial use requires “Built with MiniMax M3.1-Flash” attribution and a notice to, or authorization from, MiniMax. See the license for the terms and its prohibited uses.

Elsewhere

• GGUF files, method and full evaluation: orcarouter/MiniMax-M3.1-Flash-abliterated-GGUF

One Gateway. Every Model.

Route Smarter · Ship Safer · Spend Less

Website · Models · GitHub · Discord · X