Instructions to use mudler/flint-APEX-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 mudler/flint-APEX-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 mudler/flint-APEX-GGUF
# Run inference directly in the terminal:
llama cli -hf mudler/flint-APEX-GGUF

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf mudler/flint-APEX-GGUF
# Run inference directly in the terminal:
llama cli -hf mudler/flint-APEX-GGUF

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 mudler/flint-APEX-GGUF
# Run inference directly in the terminal:
./llama-cli -hf mudler/flint-APEX-GGUF

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 mudler/flint-APEX-GGUF
# Run inference directly in the terminal:
./build/bin/llama-cli -hf mudler/flint-APEX-GGUF

Use Docker

docker model run hf.co/mudler/flint-APEX-GGUF

Llama new
LM Studio
Jan
Ollama
How to use mudler/flint-APEX-GGUF with Ollama:
```
ollama run hf.co/mudler/flint-APEX-GGUF
```
Unsloth Desktop

How to use mudler/flint-APEX-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf mudler/flint-APEX-GGUF

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": "mudler/flint-APEX-GGUF"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use mudler/flint-APEX-GGUF with Docker Model Runner:
```
docker model run hf.co/mudler/flint-APEX-GGUF
```

Lemonade

How to use mudler/flint-APEX-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull mudler/flint-APEX-GGUF

Run and chat with the model

lemonade run user.flint-APEX-GGUF-{{QUANT_TAG}}

List all available models

lemonade list

Hermes Agent

How to use mudler/flint-APEX-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 mudler/flint-APEX-GGUF

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 mudler/flint-APEX-GGUF

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use mudler/flint-APEX-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf mudler/flint-APEX-GGUF

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 "mudler/flint-APEX-GGUF" \
  --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"

Flint APEX GGUF

APEX (Adaptive Precision for EXpert Models) quantizations of Flint, a 39.8B-total, about 2.5B-active mixture-of-experts model for coding agents. The base GGUF is unsloth/flint-GGUF, from RecursiveSI/flint.

Brought to you by the LocalAI team | APEX Project | Technical Report

Benchmark Results

Perplexity is measured on wikitext-2-raw with context 2048. The BF16 reference perplexity is 22.87 (+/- 0.25). HellaSwag and Winogrande use 400 tasks each, zero-shot. The numbers were measured on the files as built on a GPU host, using llama.cpp at commit a37cc461.

Model	Size (GiB)	Perplexity	HellaSwag	Winogrande
APEX I-Quality	23.2	23.14	72.75	67.0
APEX Quality	23.2	23.25	72.75	66.0
APEX I-Balanced	26.6	23.24	73.0	67.0
APEX Balanced	26.6	22.82	72.75	67.5
APEX I-Compact	17.2	22.62	73.5	66.5
APEX Compact	17.2	22.83	73.25	67.0
APEX I-Mini	13.8	24.81	71.0	65.25
BF16 reference	—	22.87	—	—

Only these benchmarks were run. No MMLU, ARC or TruthfulQA numbers are claimed.

What is APEX?

APEX is a quantization strategy for Mixture-of-Experts (MoE) models. It classifies tensors by role (routed expert, shared expert, attention) and applies a layer-wise precision gradient: edge layers get higher precision, middle layers get more aggressive compression. The I-variants use imatrix-guided quantization.

See the APEX project for details and the technical report.

Files

File	Size
Flint-APEX-I-Quality.gguf	23.2 GiB
Flint-APEX-Quality.gguf	23.2 GiB
Flint-APEX-I-Balanced.gguf	26.6 GiB
Flint-APEX-Balanced.gguf	26.6 GiB
Flint-APEX-I-Compact.gguf	17.2 GiB
Flint-APEX-Compact.gguf	17.2 GiB
Flint-APEX-I-Mini.gguf	13.8 GiB

Architecture

Model: Flint (flint architecture) by Recursive
Layers: 49, with a sliding-window attention pattern (window 513) and full attention on the remaining layers
Experts: 256 routed, 6 active per token, plus 2 shared experts
Context: up to 262,144 tokens
Total parameters: about 39.8B
Active parameters: about 2.5B per token

This build needs a llama.cpp that includes Flint support (llama.cpp commit a37cc461 or later).

Run with LocalAI

local-ai run mudler/flint-APEX-GGUF@Flint-APEX-I-Balanced.gguf

Credits

APEX is brought to you by the LocalAI team. Developed through human-driven, AI-assisted research. Built on llama.cpp.

Downloads last month: 1,894

GGUF

Model size

40B params

Architecture

flint

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for mudler/flint-APEX-GGUF

Base model

RecursiveSI/flint

Quantized

unsloth/flint-GGUF

Quantized

(1)

this model

Collection including mudler/flint-APEX-GGUF

APEX Quants (GGUF)

Collection

MoE models quantized with the APEX Quantization technique ( https://github.com/mudler/apex-quant ) • 50 items • Updated 2 days ago • 153