Instructions to use batiai/GLM-5.1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use batiai/GLM-5.1-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="batiai/GLM-5.1-GGUF",
	filename="zai-org-GLM-5.1-IQ3_XXS-00001-of-00007.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use batiai/GLM-5.1-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS
# Run inference directly in the terminal:
llama-cli -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS
# Run inference directly in the terminal:
llama-cli -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS
# Run inference directly in the terminal:
./llama-cli -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS
# Run inference directly in the terminal:
./build/bin/llama-cli -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Use Docker

docker model run hf.co/batiai/GLM-5.1-GGUF:IQ3_XXS

LM Studio
Jan

vLLM

How to use batiai/GLM-5.1-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "batiai/GLM-5.1-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "batiai/GLM-5.1-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/batiai/GLM-5.1-GGUF:IQ3_XXS

Ollama
How to use batiai/GLM-5.1-GGUF with Ollama:
```
ollama run hf.co/batiai/GLM-5.1-GGUF:IQ3_XXS
```

Unsloth Studio new

How to use batiai/GLM-5.1-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for batiai/GLM-5.1-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for batiai/GLM-5.1-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for batiai/GLM-5.1-GGUF to start chatting

Pi new

How to use batiai/GLM-5.1-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "batiai/GLM-5.1-GGUF:IQ3_XXS"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use batiai/GLM-5.1-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf batiai/GLM-5.1-GGUF:IQ3_XXS

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default batiai/GLM-5.1-GGUF:IQ3_XXS

Run Hermes

hermes

Docker Model Runner
How to use batiai/GLM-5.1-GGUF with Docker Model Runner:
```
docker model run hf.co/batiai/GLM-5.1-GGUF:IQ3_XXS
```

Lemonade

How to use batiai/GLM-5.1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull batiai/GLM-5.1-GGUF:IQ3_XXS

Run and chat with the model

lemonade run user.GLM-5.1-GGUF-IQ3_XXS

List all available models

lemonade list

GLM-5.1 GGUF — Quantized by BatiAI

IQ3_XXS / IQ4_XS quantization of zai-org/GLM-5.1 (744B total / 40B active MoE). Quantized directly from official Z.AI weights by BatiAI.

Why GLM-5.1?

744B parameters (40B active) — frontier MoE with Deep Sparse Attention (DSA)
#1 open-source on SWE-Bench Pro — leads the open-weight pack on agentic coding
256 experts per layer (top-8 routing + DSA indexer) — extreme sparsity
79 transformer blocks with hybrid attention/FFN routing
MIT license — fully permissive for commercial use, fine-tuning, redistribution
Released by Z.AI / Zhipu AI — same lineage as ChatGLM / GLM-4

Quick Start

# IQ3_XXS (smaller, 273GB — needs 320GB+ unified RAM)
hf download batiai/GLM-5.1-GGUF --include "*IQ3_XXS*"

# IQ4_XS (recommended balance, 376GB — needs 448GB+ unified RAM)
hf download batiai/GLM-5.1-GGUF --include "*IQ4_XS*"

Available Quantizations

Quant	Total Size	Shards	Min RAM	Target Hardware
IQ3_XXS	273 GB	7 × ~40 GB	~320 GB	M3 Ultra 512GB / H100 node
IQ4_XS	376 GB	9 × ~42 GB	~448 GB	M3 Ultra 512GB / 8× A100 80GB

⚠️ Not for consumer Mac — workstation / server class. 16–128GB Macs should use batiai/qwen3.6-35b or batiai/minimax-m2.7. Mac Studio M2 Ultra 192GB users should use batiai/kimi-k2.6:iq3 (394GB but lighter active MoE) — GLM-5.1 is denser at 40B active.

Hardware Reality Check

Your System	IQ3_XXS (273GB)	IQ4_XS (376GB)
Mac 128GB	❌ Won't fit	❌
Mac 192GB	⚠️ Heavy swap (unusable)	❌
Mac 256GB	⚠️ Tight (~50GB swap)	❌
Mac 384GB	✅ Usable	⚠️ Tight
Mac M3 Ultra 512GB	✅ Comfortable	✅ Usable
2× M3 Ultra (cluster)	✅ Fast	✅ Fast
8× A100 80GB (640GB)	✅ Fast	✅ Fast
H100 node	✅ Fast	✅ Fast

Numbers based on MoE activation pattern — 40B active params × 2 bytes (Q4 active) ≈ 80GB runtime, plus shard buffers + KV cache (32K ctx ≈ 8-12GB). Going below the min RAM forces SSD paging which destroys throughput.

Special Engineering Notes

GLM-5.1 uses Deep Sparse Attention (DSA) — a per-layer "indexer" tensor selects the top-K key positions for sparse attention. This required two fixes during quantization:

DSA indexer tensors not in imatrix — --tensor-type indexer=q5_k override (~600 MB overhead total)
Last block (blk.78) imatrix gap — bati.cpp llama-imatrix does not record the final block; --tensor-type blk.78=q5_k workaround applied

Both flags are baked into our quantization pipeline (scripts/runtime/glm-pipeline.sh). The fallback Q5_K layer adds < 0.2% to file size but prevents low-bit IQ-quants from bailing on missing imatrix data.

79 of 1809 tensors used fallback quantization — these are the indexer + last-block weights kept at higher precision.

What BatiAI's Quantization Delivers

	BatiAI	typical 3rd-party
Source	Direct from official Z.AI weights	Often re-quantized from other GGUFs
Quantization flow	safetensors → Q8_0 → IQ3_XXS/IQ4_XS with imatrix (wikitext-2, 200 chunks)	Varies
imatrix	✅ 200 chunks (quality saturation)	Often skipped or fewer chunks
DSA indexer handling	✅ Q5_K override documented	Often unaddressed → garbage low-bit
Last-block imatrix gap	✅ Workaround applied	Often causes bail-out or quality loss
BatiAI signature	✅ `general.author=BatiAI`, `general.url=https://flow.bati.ai`	✗

Model Comparison — BatiAI Lineup

Your Hardware	Best BatiAI Model	Size
16GB Mac	`batiai/gemma4-e4b:q4`	5GB
24GB Mac	`batiai/gemma4-26b:iq4`	15GB
48GB Mac	`batiai/qwen3.6-35b:iq4`	22GB
96GB Mac	`batiai/qwen3.6-35b:q6`	29GB
128GB Mac	`batiai/minimax-m2.7:iq3`	82GB
192GB Mac Studio	`batiai/kimi-k2.6:iq3`	394GB (paged)
M3 Ultra 512GB	`batiai/GLM-5.1:iq4` ⬅ here	376GB
M3 Ultra 512GB (alt)	`batiai/kimi-k2.6:iq4`	546GB (heavy swap)

GLM-5.1 IQ4_XS at 376 GB is the largest model that runs comfortably on a single M3 Ultra 512GB without SSD swap. Kimi K2.6 IQ4 (546GB) would page heavily on the same machine.

Benchmarks (source model)

Benchmark	GLM-5.1	Notes
SWE-Bench Pro	#1 open-source	Beats Kimi K2.6 (58.6) on coding tasks
HumanEval	High	Strong code generation
MMLU	Strong	General reasoning
Context	32K (extendable via YARN)
Tool use	✅ Native	Function calling supported

Numbers are from Z.AI's official report. Validating quantization preserves these on Mac M3 Ultra is pending (bench.sh on target hardware).

Technical Details

Original Model: zai-org/GLM-5.1
Architecture: GlmMoeDsaForCausalLM — 744B total / 40B active, 79 blocks (3 dense + 76 MoE), 256 experts (top-8), DSA hybrid attention
Original storage: BF16/FP8 mix (~1.4 TB safetensors)
License: MIT
Quantized with: bati.cpp v0.1.2 (BatiAI's llama.cpp fork — needed for DSA architecture)
Calibration: wikitext-2-raw, 200 chunks (quality saturation)
imatrix overrides: --tensor-type indexer=q5_k --tensor-type blk.78=q5_k
Quantized by: BatiAI

Usage

llama.cpp / bati.cpp

GLM-5.1 currently requires bati.cpp (BatiAI's llama.cpp fork) — mainline ggml-org/llama.cpp does not yet support glm-dsa architecture. Will switch to mainline once support lands.

git clone https://github.com/batiai/bati.cpp.git
cd bati.cpp
cmake -B build -DGGML_METAL=ON   # macOS
# or: cmake -B build -DGGML_CUDA=ON   # Linux
cmake --build build -j --target llama-cli

hf download batiai/GLM-5.1-GGUF --include "*IQ4_XS*" --local-dir ./glm51

build/bin/llama-cli -m ./glm51/zai-org-GLM-5.1-IQ4_XS-00001-of-00009.gguf \
    -p "Your prompt" \
    --ctx-size 32768 \
    --n-gpu-layers 99

Ollama

Ollama support pending — will require glm-dsa arch upstream in ggml-org/llama.cpp first.

vLLM / TGI

Not directly compatible — these serve FP8/BF16 safetensors. Use original zai-org/GLM-5.1 for vLLM.

About `bati.cpp`

batiai/bati.cpp is BatiAI's llama.cpp-based fork focused on:

Apple Silicon (Metal) optimization
Frontier-model early access (V4-Flash, GLM-5.1 DSA, etc.) before mainline merges
BatiAI quantization standard (signature, imatrix workflow)

Built on top of ggml-org/llama.cpp and antirez/llama.cpp-deepseek-v4-flash (all MIT). See bati.cpp/ATTRIBUTION.md for full credits.

License

Inherits the source model license: MIT.

About BatiFlow

BatiFlow — free on-device AI automation for Mac. 5MB native app, 60+ tools (KakaoTalk, iMessage, Slack, Calendar, Notes, Chrome, file system). Works with all batiai/* models.

Downloads last month: 1,493

GGUF

Model size

754B params

Architecture

glm-dsa

Hardware compatibility

3-bit

4-bit

Model tree for batiai/GLM-5.1-GGUF

Base model

zai-org/GLM-5.1

Quantized

(40)

this model

Collection including batiai/GLM-5.1-GGUF

🚀 Frontier MoE — 128B–1T

Collection

Largest open-weight LLMs, BatiAI-quantized. Mac-runnable from M4 Max 128GB to Mac Studio M3 Ultra 512GB. • 7 items • Updated 6 days ago