Instructions to use mattjoyce/antares-1b-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 mattjoyce/antares-1b-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 mattjoyce/antares-1b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf mattjoyce/antares-1b-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 mattjoyce/antares-1b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf mattjoyce/antares-1b-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 mattjoyce/antares-1b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf mattjoyce/antares-1b-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 mattjoyce/antares-1b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf mattjoyce/antares-1b-GGUF:Q4_K_M

Use Docker

docker model run hf.co/mattjoyce/antares-1b-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use mattjoyce/antares-1b-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mattjoyce/antares-1b-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": "mattjoyce/antares-1b-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mattjoyce/antares-1b-GGUF:Q4_K_M

Ollama
How to use mattjoyce/antares-1b-GGUF with Ollama:
```
ollama run hf.co/mattjoyce/antares-1b-GGUF:Q4_K_M
```
Unsloth Desktop

How to use mattjoyce/antares-1b-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf mattjoyce/antares-1b-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "mattjoyce/antares-1b-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use mattjoyce/antares-1b-GGUF with Docker Model Runner:
```
docker model run hf.co/mattjoyce/antares-1b-GGUF:Q4_K_M
```

Lemonade

How to use mattjoyce/antares-1b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull mattjoyce/antares-1b-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.antares-1b-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use mattjoyce/antares-1b-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf mattjoyce/antares-1b-GGUF:Q4_K_M

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 mattjoyce/antares-1b-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use mattjoyce/antares-1b-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf mattjoyce/antares-1b-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "mattjoyce/antares-1b-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

mattjoyce commited on Aug 1

Commit

4d15386

verified ·

1 Parent(s): c7fd470

card: charts embedded, phrasing scrubbed

Browse files

Files changed (1) hide show

README.md +13 -9

README.md CHANGED Viewed

@@ -24,12 +24,14 @@ Nine quantizations of [fdtn-ai/antares-1b](https://huggingface.co/fdtn-ai/antare
 through a terminal and submits a ranked list of suspect files; a fine-tune of
 [ibm-granite/granite-4.0-1b](https://huggingface.co/ibm-granite/granite-4.0-1b)).
-What's different about this repo: **every file was benchmarked before publishing** on
-the model's own vulnerability-localization benchmark — the full 499-task set,
-detection (Phase A) *and* clean-tree false-alarm rate (Phase B), with three-run error
-bars at the decision boundaries — and every file's sha256 is published. As of
-2026-08-02, none of the dozen-plus community GGUF repos of this model we surveyed
-publish per-file hashes or measurements, and we found no imatrix builds among them.
 ## Which file do I want?
@@ -58,7 +60,7 @@ rate on clean trees. Rungs 0–6 measured 2026-07-30/31; rungs 7–8 measured 20
 | 5 | Q3_K_M | — | 0.95 GB | 26% | 0.0402 | 51.3% | 0.539* |
 | 6 | Q2_K | — | 0.77 GB | 21% | 0.0000 | 100% | 1.000* |
-**Reading this table honestly:**
 - ± values are 1 sd over three full benchmark runs. Cells without ± are single runs;
   observed single-run spread at a fixed rung is roughly ±0.003–0.006, so **rungs are
@@ -69,6 +71,8 @@ rate on clean trees. Rungs 0–6 measured 2026-07-30/31; rungs 7–8 measured 20
 - **IQ4_XS**: three runs scored 0.1742 / 0.1588 / 0.1613 on identical config. With
   n=3 the wide spread is suggestive, not proven — but treat any single-run IQ4_XS
   benchmark (ours or anyone's) as uninformative.
 - **None of these rungs is a stand-alone merge gate.** Healthy rungs flag something
   on 96–99% of *clean* trees (Phase B). Treat output as leads for a human or a
   stronger model — which matches the upstream guidance ("a lead to verify, not
@@ -104,7 +108,7 @@ Measurement configuration (what these numbers were produced with):
 Traps:
 1. **Ollama's `/v1/completions` is not raw.** It wraps prompts in the GGUF's chat
-   template with no error and quietly degraded results. If you must use Ollama, set an
    identity template (`TEMPLATE {{ .Prompt }}`) and verify with a token-count sentinel
    (tokenize your prompt separately; compare `prompt_tokens`).
 2. **Double-BOS.** The tokenizer adds BOS; if your client also prepends one, quality
@@ -177,4 +181,4 @@ Apache-2.0, inherited from [fdtn-ai/antares-1b](https://huggingface.co/fdtn-ai/a
 is gated — if you want the original safetensors, the CLI tooling, or the benchmark,
 go through their gate; this repo exists to publish *measured* quantizations, not to
 route around upstream. If you are the upstream team and want anything here changed,
-open a discussion — happy to adjust.

 through a terminal and submits a ranked list of suspect files; a fine-tune of
 [ibm-granite/granite-4.0-1b](https://huggingface.co/ibm-granite/granite-4.0-1b)).
+**Every file here was benchmarked before publishing** on the model's own
+vulnerability-localization benchmark — the full 499-task set, detection (Phase A)
+*and* clean-tree false-alarm rate (Phase B), with three-run error bars at the
+decision boundaries — and every file's sha256 is published. As of 2026-08-02, none
+of the dozen-plus community GGUF repos of this model we surveyed publish per-file
+hashes or measurements, and we found no imatrix builds among them.
+![Quant ladder: File F1 vs file size](ladder.svg)
 ## Which file do I want?
 | 5 | Q3_K_M | — | 0.95 GB | 26% | 0.0402 | 51.3% | 0.539* |
 | 6 | Q2_K | — | 0.77 GB | 21% | 0.0000 | 100% | 1.000* |
+**Table notes:**
 - ± values are 1 sd over three full benchmark runs. Cells without ± are single runs;
   observed single-run spread at a fixed rung is roughly ±0.003–0.006, so **rungs are
 - **IQ4_XS**: three runs scored 0.1742 / 0.1588 / 0.1613 on identical config. With
   n=3 the wide spread is suggestive, not proven — but treat any single-run IQ4_XS
   benchmark (ours or anyone's) as uninformative.
+![Run-to-run spread by rung](stability.svg)
 - **None of these rungs is a stand-alone merge gate.** Healthy rungs flag something
   on 96–99% of *clean* trees (Phase B). Treat output as leads for a human or a
   stronger model — which matches the upstream guidance ("a lead to verify, not
 Traps:
 1. **Ollama's `/v1/completions` is not raw.** It wraps prompts in the GGUF's chat
+   template — no error, degraded results. If you must use Ollama, set an
    identity template (`TEMPLATE {{ .Prompt }}`) and verify with a token-count sentinel
    (tokenize your prompt separately; compare `prompt_tokens`).
 2. **Double-BOS.** The tokenizer adds BOS; if your client also prepends one, quality
 is gated — if you want the original safetensors, the CLI tooling, or the benchmark,
 go through their gate; this repo exists to publish *measured* quantizations, not to
 route around upstream. If you are the upstream team and want anything here changed,
+open a discussion.