Instructions to use pinkelephantlimited/phone-helper-app with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use pinkelephantlimited/phone-helper-app 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 pinkelephantlimited/phone-helper-app:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pinkelephantlimited/phone-helper-app:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf pinkelephantlimited/phone-helper-app:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pinkelephantlimited/phone-helper-app: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 pinkelephantlimited/phone-helper-app:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf pinkelephantlimited/phone-helper-app: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 pinkelephantlimited/phone-helper-app:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf pinkelephantlimited/phone-helper-app:Q4_K_M

Use Docker

docker model run hf.co/pinkelephantlimited/phone-helper-app:Q4_K_M

LM Studio
Jan
Ollama
How to use pinkelephantlimited/phone-helper-app with Ollama:
```
ollama run hf.co/pinkelephantlimited/phone-helper-app:Q4_K_M
```

Unsloth Studio

How to use pinkelephantlimited/phone-helper-app 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 pinkelephantlimited/phone-helper-app 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 pinkelephantlimited/phone-helper-app to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for pinkelephantlimited/phone-helper-app to start chatting

How to use pinkelephantlimited/phone-helper-app with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf pinkelephantlimited/phone-helper-app:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "pinkelephantlimited/phone-helper-app:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use pinkelephantlimited/phone-helper-app with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf pinkelephantlimited/phone-helper-app: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 "pinkelephantlimited/phone-helper-app: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"

Docker Model Runner
How to use pinkelephantlimited/phone-helper-app with Docker Model Runner:
```
docker model run hf.co/pinkelephantlimited/phone-helper-app:Q4_K_M
```

Lemonade

How to use pinkelephantlimited/phone-helper-app with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull pinkelephantlimited/phone-helper-app:Q4_K_M

Run and chat with the model

lemonade run user.phone-helper-app-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use pinkelephantlimited/phone-helper-app with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf pinkelephantlimited/phone-helper-app: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 pinkelephantlimited/phone-helper-app:Q4_K_M

Run Hermes

hermes

Atomic Chat

Pink Elephant Talk — v7.1

Pink Elephant Talk is the flagship mobile product of Pink Elephant Limited (Hong Kong) — a commercial AI technology development company.

It is a fully offline, on-device chat assistant with photo vision. The vision-language model runs entirely on your phone: no cloud, no account, no API fees, no telemetry. Your conversations, photos and documents never leave the device.

📥 Download the APK

Latest release — Pink Elephant Talk v7.1 (voice output):

Download PinkElephantTalk-v7.1.apk (157.9 MB · SHA-256 a863092b3731f76648d09e31ef352613eec353d3e19859ac316da20f7a4d0c54)

All versions are listed in the Files and versions tab of this repository, and in the table below.

v7.1 Highlights

Voice output (text-to-speech) — tap the 🔊 button in the top-right to read every assistant reply aloud automatically, or tap 🔊 under any assistant bubble to speak / ⏹ stop that message. Uses the phone's native speech engine, fully on-device.

v7.0 Highlights

Photo vision — attach a photo with the 🖼️ button; the on-device model sees and describes it (verified on hardware with the pink-elephant logo).
Core model: Qwen3-VL-2B-Thinking (Apache-2.0) — a phone-class vision-language model, ~1.5 GB one-time download (language model + vision encoder), cached on device.
Visible thinking mode — the model shows its reasoning, then gives a clear final answer.
Reliable web search — a client-side topic detector fetches fresh data for weather, news, prices and sports even when the model answers from memory. Weather comes from Open-Meteo (no API key, CC BY 4.0), headlines from Google News RSS, summaries from Wikipedia; sources are cited.
The model has the right to the internet — it never claims it "cannot access the internet"; live questions (e.g. HK news, weather) always search.
Document upload — attach a .txt, .pdf or .docx; text is extracted on-device (PDF via pdfbox-android) and the model can answer about it.
Branded identity — answers "I was created by Pink Elephant Limited."
On-device, offline — works with zero connectivity after first setup. Your chats, photos and documents stay on the phone.

Releases

APK (click to download)	Model	Notes
`PinkElephantTalk-v7.1.apk`	Qwen3-VL-2B-Thinking (Q4_K_M + mmproj)	Voice output (TTS), photo vision, thinking mode, web search, document upload (current)
`PinkElephantTalk-v7.0.apk`	Qwen3-VL-2B-Thinking (Q4_K_M + mmproj)	Photo vision, thinking mode, web search, document upload
`PinkElephantTalk-v6.2.apk`	Qwen3-1.7B-Instruct (Q4_K_M)	Internet-right prompt + duplicate-answer fix
`PinkElephantTalk-v6.1.apk`	Qwen3-1.7B-Instruct (Q4_K_M)	Added document upload + thumbnail logo
`PinkElephantTalk-v6.0.apk`	Qwen3-1.7B-Instruct (Q4_K_M)	Reliable web search + branding
`PinkElephantTalk-v5.0.apk`	Qwen3-1.7B-Instruct (Q4_K_M)	Thinking mode (visible reasoning)
`PinkElephantTalk-v4.0.apk`	Qwen2.5-3B-Instruct (Q4_K_M)	Text chat, offline, rose UI
`PinkElephantTalk-v3.0.1.apk`	Qwen2-VL-2B-Instruct (Q4_0)	Vision+voice assistant (legacy)
`PinkElephantTalk-v3.0.apk`	Qwen2-VL-2B-Instruct (Q4_0)	Vision+voice assistant (legacy)
`PinkElephantTalk-v2.0.apk`	Qwen2.5-VL-3B-Instruct	Vision (legacy)
`PinkElephantTalk-v1.0.apk`	Qwen2.5-VL-3B-Instruct	Initial release (legacy)

Install

Download PinkElephantTalk-v7.0.apk.
Allow installs from unknown sources (Settings → Security).
Open the app. On first launch it downloads the model (~1.5 GB, one-time, with per-file progress labels), then runs entirely offline.

On some phones (e.g. vivo) a security check appears on first install; tap the checkbox and continue — the app requests no sensitive permissions.

Model

GGUF: Qwen/Qwen3-VL-2B-Thinking-GGUF — Qwen3VL-2B-Thinking-Q4_K_M.gguf (1,107,409,888 bytes) + vision projector mmproj-Qwen3VL-2B-Thinking-Q8_0.gguf (445,053,216 bytes).
Vision config: 1024 image tokens (minimum/maximum).
Inference engine: llama.cpp via llama.rn (both MIT).

Commercial documents

The complete commercial package is included in this repository under docs/commercial/:

Document	File
Company profile	`docs/commercial/company-profile.md`
Product line brochure	`docs/commercial/product-line-brochure.md`
Data sheet (v7.0)	`docs/commercial/data-sheet.md`
Pricing & monetization plan	`docs/commercial/pricing-plan.md`
Privacy policy	`docs/commercial/privacy-policy.md`
Terms of service	`docs/commercial/terms-of-service.md`
Release notes (v1.0 → v7.0)	`docs/commercial/release-notes.md`
Partnership & distribution prospectus	`docs/commercial/partnership-prospectus.md`

Compliance & Licensing

Base model: redistributed under Apache-2.0 by preserving the Qwen copyright/license notice — see THIRD-PARTY-NOTICES in the source repo. "Qwen" is a trademark of Alibaba Group.
Web data: Open-Meteo (CC BY 4.0), Google News RSS, Wikipedia (CC BY-SA). PDF text extraction via pdfbox-android (Apache-2.0).
No third-party cloud inference: Pink Elephant Talk performs inference on-device; it does not call OpenAI, Google, or any hosted LLM API.

About Pink Elephant Limited

An official commercial AI technology development company registered in Hong Kong, building a family of private, on-device mobile products.

Downloads last month: -

GGUF

Model size

2B params

Architecture

qwen3vl

Hardware compatibility

4-bit

Inference Providers NEW

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