Instructions to use anthonym21/slipstream-glm-z1-9b-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries
PEFT
How to use anthonym21/slipstream-glm-z1-9b-gguf with PEFT:
```
Task type is invalid.
```

How to use anthonym21/slipstream-glm-z1-9b-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="anthonym21/slipstream-glm-z1-9b-gguf",
	filename="slipstream-Q4_K_M.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 anthonym21/slipstream-glm-z1-9b-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

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 anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

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 anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

Use Docker

docker model run hf.co/anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use anthonym21/slipstream-glm-z1-9b-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "anthonym21/slipstream-glm-z1-9b-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": "anthonym21/slipstream-glm-z1-9b-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

Ollama
How to use anthonym21/slipstream-glm-z1-9b-gguf with Ollama:
```
ollama run hf.co/anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
```

Unsloth Studio new

How to use anthonym21/slipstream-glm-z1-9b-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 anthonym21/slipstream-glm-z1-9b-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 anthonym21/slipstream-glm-z1-9b-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for anthonym21/slipstream-glm-z1-9b-gguf to start chatting

Docker Model Runner
How to use anthonym21/slipstream-glm-z1-9b-gguf with Docker Model Runner:
```
docker model run hf.co/anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M
```

Lemonade

How to use anthonym21/slipstream-glm-z1-9b-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull anthonym21/slipstream-glm-z1-9b-gguf:Q4_K_M

Run and chat with the model

lemonade run user.slipstream-glm-z1-9b-gguf-Q4_K_M

List all available models

lemonade list

Slipstream GLM-Z1-9B

A finetuned version of GLM-Z1-9B-0414 trained on the Slipstream protocol - a semantic quantization system that achieves 82% token reduction in multi-agent AI communication.

Model Description

This model has learned the Think-Quantize-Transmit (TQT) cognitive pattern:

THINK: Reason about the communication intent
QUANTIZE: Map intent to a semantic anchor in the UCR manifold
TRANSMIT: Output a compact SLIP wire format message

Example

Input:

Tell bob to review my authentication code

Output:

THOUGHT: I need bob to do a code review on the auth module
QUANTIZE: [ACTION=request | DOMAIN=task | URGENCY=normal | POLARITY=neutral] -> RequestReview
SLIP: SLIP v1 alice bob RequestReview auth_module

Training Details

Parameter	Value
Base Model	zai-org/GLM-Z1-9B-0414
Method	LoRA (rank=16, alpha=16)
Epochs	2
Learning Rate	2e-4
Batch Size	16 (4 × 4 grad accum)
Sequence Length	2048
Training Examples	2,283
Hardware	Google Colab (A100/V100)
Framework	Unsloth + TRL

LoRA Target Modules

Attention: q_proj, k_proj, v_proj, o_proj
MLP: gate_proj, up_proj, down_proj

Available Formats

Format	Repository	Use Case
LoRA Adapter	slipstream-glm-z1-9b	Merge with base model
Merged 16-bit	slipstream-glm-z1-9b-merged	Direct loading
GGUF Q4_K_M	slipstream-glm-z1-9b-gguf	Ollama / llama.cpp
GGUF Q8_0	slipstream-glm-z1-9b-gguf	Higher quality local

Usage

With Transformers + PEFT

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = AutoModelForCausalLM.from_pretrained("zai-org/GLM-Z1-9B-0414")
model = PeftModel.from_pretrained(base_model, "anthonym21/slipstream-glm-z1-9b")
tokenizer = AutoTokenizer.from_pretrained("anthonym21/slipstream-glm-z1-9b")

With Ollama

# Download GGUF
wget https://huggingface.co/anthonym21/slipstream-glm-z1-9b-gguf/resolve/main/slipstream-q4_k_m.gguf

# Create Modelfile
cat > Modelfile <<EOF
FROM ./slipstream-q4_k_m.gguf
SYSTEM "You are an AI agent using the Slipstream protocol for efficient multi-agent communication."
EOF

# Run
ollama create slipstream -f Modelfile
ollama run slipstream "Tell bob to review my code"

With Unsloth (for inference)

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    "anthonym21/slipstream-glm-z1-9b",
    max_seq_length=2048,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)

UCR Anchors

The model understands 21 core anchors:

Category	Anchors
Requests	`RequestTask`, `RequestReview`, `RequestHelp`, `RequestPlan`
Inform	`InformComplete`, `InformProgress`, `InformBlocked`, `InformStatus`
Propose	`ProposePlan`, `ProposeChange`, `ProposeAlternative`
Evaluate	`EvalApprove`, `EvalReject`, `EvalNeedsWork`
Meta	`Accept`, `Reject`, `MetaAck`, `MetaHandoff`, `Fallback`

Wire Format

SLIP v1 <src> <dst> <anchor> [payload...]

Example: SLIP v1 alice bob RequestReview auth_module

Related Resources

Protocol Spec: github.com/anthony-maio/slipcore
Training Dataset: hf.co/anthonym21/slipstream-tqt
Paper: Slipstream: Semantic Quantization for Efficient Multi-Agent Coordination
PyPI: pip install slipcore

Citation

@misc{maio2025slipstream,
  title={Slipstream: Semantic Quantization for Efficient Multi-Agent Coordination},
  author={Maio, Anthony},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/anthonym21/slipstream-glm-z1-9b}
}

License

Apache 2.0

Downloads last month: 15

GGUF

Model size

9B params

Architecture

glm4

Hardware compatibility

4-bit

8-bit

Model tree for anthonym21/slipstream-glm-z1-9b-gguf

Base model

zai-org/GLM-Z1-9B-0414

Adapter

(5)

this model

Dataset used to train anthonym21/slipstream-glm-z1-9b-gguf

Collection including anthonym21/slipstream-glm-z1-9b-gguf

Streamlined Inter-agent Protocol (Slipstream)

Collection

Semantic Quantization for Efficient Multi-Agent Coordination • 4 items • Updated Mar 2 • 1