Instructions to use QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M

Use Docker

docker model run hf.co/QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use QuantPasture/Step-3.5-Flash-GGUF with Ollama:
```
ollama run hf.co/QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
```
Unsloth Desktop

How to use QuantPasture/Step-3.5-Flash-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf QuantPasture/Step-3.5-Flash-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": "QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use QuantPasture/Step-3.5-Flash-GGUF with Docker Model Runner:
```
docker model run hf.co/QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M
```

Lemonade

How to use QuantPasture/Step-3.5-Flash-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Step-3.5-Flash-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-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 QuantPasture/Step-3.5-Flash-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use QuantPasture/Step-3.5-Flash-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf QuantPasture/Step-3.5-Flash-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 "QuantPasture/Step-3.5-Flash-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"

AesSedai commited on Jun 3

Commit

c7bb172

verified ·

1 Parent(s): 2e31c87

Update README.md

Browse files

Files changed (1) hide show

README.md +11 -5

README.md CHANGED Viewed

@@ -3,13 +3,19 @@ base_model:
 - stepfun-ai/Step-3.5-Flash
 ---
 This repo contains specialized MoE-quants for Step-3.5-Flash. The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization. To that end, the quantization type default is kept in high quality and the FFN UP + FFN GATE tensors are quanted down along with the FFN DOWN tensors.
-| Quant | Size | Mixture | PPL | 1-(Mean PPL(Q)/PPL(base)) | KLD |
-| :--------- | :--------- | :------- | :------- | :------- | :------- |
-| Q4_K_M | 113.82 GiB (4.96 BPW) | Q8_0 / Q4_K / Q4_K / Q5_K | 4.718049 ± 0.030373 | +0.3762% | 0.015464 ± 0.000133 |
-| IQ4_XS | 88.90 GiB (3.88 BPW) | Q8_0 / IQ3_S / IQ3_S / IQ4_XS | 4.822499 ± 0.031236 | +2.5984% | 0.042753 ± 0.000301 |
-| IQ3_XXS | 73.10 GiB (3.19 BPW) | Q6_K / IQ3_XXS / IQ3_XXS / IQ3_XXS | 4.882908 ± 0.031560 | +3.8836% | 0.078681 ± 0.000506 |
 ![kld_graph](kld_data/01_kld_vs_filesize.png "Chart showing Pareto KLD analysis of quants")
 ![ppl_graph](kld_data/02_ppl_vs_filesize.png "Chart showing Pareto PPL analysis of quants")

 - stepfun-ai/Step-3.5-Flash
 ---
+## Updates
+- 6/3/2026: I've updated all of the quants to include the updated GGUF conversion + MTP heads (as Q8_0)
 This repo contains specialized MoE-quants for Step-3.5-Flash. The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization. To that end, the quantization type default is kept in high quality and the FFN UP + FFN GATE tensors are quanted down along with the FFN DOWN tensors.
+| Quant  | Size                  | Mixture                          | PPL                 | 1-(Mean PPL(Q)/PPL(base)) | KLD                 |
+| :----- | :-------------------- | :------------------------------- | :------------------ | :------------------------ | :------------------ |
+| Q8_0   | 197.43 GiB (8.51 BPW) | Q8_0                             | 2.419187 ± 0.011093 | +0.1197%                  | 0.005594 ± 0.000041 |
+| Q5_K_M | 138.83 GiB (5.98 BPW) | Q8_0 / Q5_K / Q5_K / Q6_K        | 2.431812 ± 0.011175 | +0.6421%                  | 0.015249 ± 0.000098 |
+| Q4_K_M | 116.22 GiB (5.01 BPW) | Q8_0 / Q4_K / Q4_K / Q5_K        | 2.477162 ± 0.011502 | +2.5190%                  | 0.040282 ± 0.000237 |
+| IQ4_XS | 91.31 GiB (3.93 BPW)  | Q8_0 / IQ3_S / IQ3_S / IQ4_XS    | 2.685042 ± 0.012863 | +11.1222%                 | 0.127994 ± 0.000696 |
+| IQ3_S  | 70.89 GiB (3.05 BPW)  | Q6_K / IQ2_S / IQ2_S / IQ3_S     | 3.313628 ± 0.017314 | +37.1367%                 | 0.355228 ± 0.001705 |
+| IQ2_S  | 64.43 GiB (2.78 BPW)  | Q6_K / IQ2_XS / IQ2_XS / IQ3_XXS | 3.772340 ± 0.020688 | +56.1208%                 | 0.488752 ± 0.002225 |
 ![kld_graph](kld_data/01_kld_vs_filesize.png "Chart showing Pareto KLD analysis of quants")
 ![ppl_graph](kld_data/02_ppl_vs_filesize.png "Chart showing Pareto PPL analysis of quants")