Instructions to use ContextReq/Pebble-25M-Chat-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use ContextReq/Pebble-25M-Chat-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ContextReq/Pebble-25M-Chat-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 ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf ContextReq/Pebble-25M-Chat-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 ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M

Use Docker

docker model run hf.co/ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use ContextReq/Pebble-25M-Chat-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ContextReq/Pebble-25M-Chat-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": "ContextReq/Pebble-25M-Chat-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M

Ollama
How to use ContextReq/Pebble-25M-Chat-GGUF with Ollama:
```
ollama run hf.co/ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
```
Unsloth Desktop
Docker Model Runner
How to use ContextReq/Pebble-25M-Chat-GGUF with Docker Model Runner:
```
docker model run hf.co/ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M
```

Lemonade

How to use ContextReq/Pebble-25M-Chat-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ContextReq/Pebble-25M-Chat-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Pebble-25M-Chat-GGUF-Q4_K_M

List all available models

lemonade list

Atomic Chat

Pebble-25M-Chat-GGUF

GGUF conversions of basically-ai/Pebble-25M-Chat (Apache 2.0).

IMPORTANT: patched llama.cpp required

Pebble uses a custom hybrid Mamba2 + attention architecture. These GGUFs carry general.architecture = "pebble", which upstream llama.cpp refuses to load.

Everything needed to run them lives in the support repo:

rootendpoint/basicallyai_llama.cpp_support

llama.cpp-pebble.patch - adds the pebble architecture to llama.cpp (applies cleanly against upstream commit 0eadefe)
basicallyai_to_gguf.py - standalone converter (numpy + safetensors only)
numpy_reference.py - independent reference implementation used to verify correctness

The chat template is embedded in the GGUF metadata; chat mode works out of the box:

llama-cli -m pebble-25m-chat-f16.gguf

Chat format: lowercase user: ... / assistant: ... turns.

Files

Quant	Size	Type
f16	49.1 MB	F16
q8_0	26.3 MB	mostly Q8_0
q4_k_m	18.5 MB	mostly Q4_K_M

Verification

Outputs were cross-checked token-by-token against an independent pure-numpy reference implementation over an 8-prompt battery (CPU and CUDA backends). Identical greedy sequences on all prompts. Quantized builds verified greedy-identical to f16 (8/8 prompts for both q8_0 and q4_k_m).

Model

25M parameters, hidden 608, 8 layers (6 Mamba2 + 2 attention), ctx 2048, vocab 2048
SFT chat variant; a research-scale model: expect toy-level output quality.

CPU support (no GPU required)

Pure-CPU support for these models (no mamba-ssm, no CUDA) lives in the basicallyai_cpu_support repository:

https://github.com/rootendpoint/basicallyai_cpu_support

It runs the original HF checkpoints in pure PyTorch on plain CPU (~118 tok/s for 10M, ~65 tok/s for 25M on a Ryzen 5 2600X), verified token-identical against this GGUF pipeline and an independent numpy oracle.