Instructions to use HuggingFaceBio/Carbon-500M-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use HuggingFaceBio/Carbon-500M-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="HuggingFaceBio/Carbon-500M-GGUF",
	filename="carbon-500m-bf16.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use HuggingFaceBio/Carbon-500M-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf HuggingFaceBio/Carbon-500M-GGUF:BF16
# Run inference directly in the terminal:
llama-cli -hf HuggingFaceBio/Carbon-500M-GGUF:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf HuggingFaceBio/Carbon-500M-GGUF:BF16
# Run inference directly in the terminal:
llama-cli -hf HuggingFaceBio/Carbon-500M-GGUF:BF16

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 HuggingFaceBio/Carbon-500M-GGUF:BF16
# Run inference directly in the terminal:
./llama-cli -hf HuggingFaceBio/Carbon-500M-GGUF:BF16

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 HuggingFaceBio/Carbon-500M-GGUF:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf HuggingFaceBio/Carbon-500M-GGUF:BF16

Use Docker

docker model run hf.co/HuggingFaceBio/Carbon-500M-GGUF:BF16

LM Studio
Jan
Ollama
How to use HuggingFaceBio/Carbon-500M-GGUF with Ollama:
```
ollama run hf.co/HuggingFaceBio/Carbon-500M-GGUF:BF16
```

Unsloth Studio new

How to use HuggingFaceBio/Carbon-500M-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 HuggingFaceBio/Carbon-500M-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 HuggingFaceBio/Carbon-500M-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for HuggingFaceBio/Carbon-500M-GGUF to start chatting

Docker Model Runner
How to use HuggingFaceBio/Carbon-500M-GGUF with Docker Model Runner:
```
docker model run hf.co/HuggingFaceBio/Carbon-500M-GGUF:BF16
```

Lemonade

How to use HuggingFaceBio/Carbon-500M-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull HuggingFaceBio/Carbon-500M-GGUF:BF16

Run and chat with the model

lemonade run user.Carbon-500M-GGUF-BF16

List all available models

lemonade list

kashif HF Staff commited on 1 day ago

Commit

673ebdf

verified ·

1 Parent(s): 4c2888c

card: HybridDNATokenizer merged upstream (#23410)

Browse files

Files changed (1) hide show

README.md +3 -3

README.md CHANGED Viewed

@@ -18,12 +18,12 @@ GGUF (bf16) conversion of [HuggingFaceBio/Carbon-500M](https://huggingface.co/Hu
 Carbon is a hybrid DNA / English language model that switches between Qwen3-4B-Base byte-level BPE for natural text and fixed 6-mer chunking for DNA inside `<dna>...</dna>` tags.
-## Requires llama.cpp with HybridDNATokenizer support
-Loading these GGUFs needs `LLAMA_VOCAB_TYPE_HYBRIDDNA`, which is not yet in upstream llama.cpp. Until the PR merges, build from the [`carbon-3b-tokenizer`](https://github.com/kashif/llama.cpp/tree/carbon-3b-tokenizer) branch:
 ```bash
-git clone -b carbon-3b-tokenizer https://github.com/kashif/llama.cpp
 cd llama.cpp && cmake -B build && cmake --build build -j
 ```

 Carbon is a hybrid DNA / English language model that switches between Qwen3-4B-Base byte-level BPE for natural text and fixed 6-mer chunking for DNA inside `<dna>...</dna>` tags.
+## Requires a recent llama.cpp
+HybridDNATokenizer support was merged in [ggml-org/llama.cpp#23410](https://github.com/ggml-org/llama.cpp/pull/23410), so any build from `master` after that works:
 ```bash
+git clone https://github.com/ggml-org/llama.cpp
 cd llama.cpp && cmake -B build && cmake --build build -j
 ```