CoolFace
Modelpublic

Youssofal/MiniMax-M2.7-GGUF

sourceHugging Faceotherupdated 6mo agoView on Hugging Face
3likes3.2kdownloads
Model Card

MiniMax-M2.7-GGUF

imatrix-calibrated GGUF quantizations of MiniMaxAI/MiniMax-M2.7 — the first model in the MiniMax lineup to participate in its own training evolution.

M2.7 matches or exceeds GPT-5.3-Codex on SWE-Pro (56.22%), rivals Opus 4.6 on VIBE-Pro (55.6%), and outperforms Claude Sonnet 4.6 on agentic benchmarks — with only 10B active parameters per token out of 229B total.

[image]

Why these quants

  • —imatrix calibrated — all sub-Q8 quants use importance matrix for optimal quality retention. No blind quantization.
  • —Q8_0 is effectively lossless — source weights are FP8 (float8_e4m3fn). Q8_0 preserves full precision.
  • —XL variants — embed and output weights kept at Q80 for meaningfully better coherence on MoE routing. Available for Q4KM, Q5KM, Q3K_M.
  • —Correct license metadata — NON-COMMERCIAL license properly set in GGUF headers via --override-kv. Other repos incorrectly list modified-MIT.
  • —Validated — every quant smoke-tested for coherent output, reasoning accuracy, and code generation before upload.

License

NON-COMMERCIAL USE ONLY. Commercial use requires prior written authorization from MiniMax. See LICENSE.

Quick Start

bash
# Recommended: Q4_K_M for most users
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q4_K_M/*" --local-dir ./

# Run with llama-server
llama-server -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 -c 32768 --jinja --reasoning-format auto -fa \
  --temp 1.0 --top-p 0.95 --top-k 40

Which file should I use?

Use CaseRecommendedSizeNotes
Maximum quality, unlimited RAMQ8_0~243 GBEffectively lossless (FP8 source)
High quality, recommendedQ6_K~188 GBNear-perfect
Best quality/size balanceQ4_K_M~138 GBDefault for most users
Better coherence at Q4Q4_K_M-XL~140 GBQ8_0 embed/output weights
Low RAM / Apple Silicon 128GBIQ4_XS~122 GBBest quality under 128 GB
Minimum viableQ3_K_M~109 GBUsable but noticeable quality loss
Absolute minimumQ2_K~83 GBSurprisingly usable

Sizing rule: Pick a quant 1-2 GB smaller than your total VRAM (GPU) or VRAM+RAM combined (Apple Silicon). A great deep-dive with charts is provided by Artefact2.

All Available Quantizations

FilenameQuantSizeSplitDescription
MiniMax-M2.7-Q8_0Q8_0~243 GBYesEffectively lossless — source is FP8
MiniMax-M2.7-Q6_KQ6_K~188 GBYesVery high quality, recommended
MiniMax-M2.7-Q5_K_MQ5KM~162 GBYesHigh quality
MiniMax-M2.7-Q5_K_SQ5KS~157 GBYesHigh quality
MiniMax-M2.7-Q4_K_M-XLQ4KM~140 GBYesQ8_0 embed/output, better coherence
MiniMax-M2.7-Q4_K_MQ4KM~138 GBYesGood quality, recommended default
MiniMax-M2.7-Q4_K_SQ4KS~130 GBYesSlightly lower quality
MiniMax-M2.7-IQ4_XSIQ4_XS~122 GBYesBest quality under 128 GB
MiniMax-M2.7-Q3_K_M-XLQ3KM~110 GBYesQ8_0 embed/output, better than standard Q3
MiniMax-M2.7-Q3_K_LQ3KL~118 GBYesLower quality, usable
MiniMax-M2.7-Q3_K_MQ3KM~109 GBYesLow quality
MiniMax-M2.7-Q3_K_SQ3KS~99 GBYesNot recommended
MiniMax-M2.7-Q2_KQ2_K~83 GBYesVery low quality, surprisingly usable

Download

bash
pip install -U "huggingface_hub[cli]"

# Most quants are split (>50 GB) -- download the folder
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q4_K_M/*" --local-dir ./

# Smaller quants (single file)
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q2_K*" --local-dir ./

Running

bash
# llama-server (recommended)
llama-server -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 -c 32768 --jinja --reasoning-format auto -fa \
  --temp 1.0 --top-p 0.95 --top-k 40

# llama-cli (conversation mode)
llama-cli -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 --jinja --reasoning-format auto -cnv

Sampling parameters (official MiniMax defaults): temperature 1.0, topp 0.95, topk 40.

Apple Silicon / MoE Notes

Despite only 10B active parameters per token, all 229B weights must reside in memory. On Apple Silicon with unified memory, use --cpu-moe for improved expert dispatch.

  • —Q4KM (~138 GB) fits on 192 GB Mac
  • —IQ4_XS (~122 GB) fits on 128 GB with swap pressure
  • —Q3KM (~109 GB) runs on 128 GB Mac

Architecture

SpecValue
Parameters229B total, 10B active (4.3% activation)
ArchitectureSparse MoE, minimax_m2
Experts256 local, 8 per token (top-k routing)
Layers62
Attention48 heads, 8 KV heads, hybrid Lightning + softmax
Context200K (rope_theta 5,000,000)
Source precisionFP8 (float8_e4m3fn)
Thinking<think>...</think> interleaved reasoning

XL Variants

XL variants (Q4KM-XL, Q5KM-XL, Q3KM-XL) keep embedding and output weight tensors at Q8_0 instead of the default lower precision. This preserves vocabulary quality at minimal size cost and meaningfully improves coherence -- particularly important for MoE models where expert routing precision matters.

Chat Template

Two templates are provided:

  • —chat_template.official.jinja -- exact copy from MiniMax, includes <think> in generation prompt
  • —chat_template.llama-server-patched.jinja -- removes <think> from generation prompt to prevent duplication in some llama-server versions

Compatibility

EngineStatus
llama.cpp (latest main)Tested
LM StudioCheck for minimax_m2 support
OllamaRequires recent version with minimax_m2
KoboldCppUntested

Quantization Details

  • —Quantized with llama.cpp at commit ff5ef82
  • —imatrix generated from bartowski calibration data
  • —License metadata corrected via --override-kv (upstream model card incorrectly propagates modified-MIT)
  • —All quants validated: coherent English output, correct reasoning, valid code generation

Credits