Cactus SDK

Cactus is the fastest cross-platform framework for deploying AI locally on smartphones.

Key Features

• Cross-Platform: Available in Flutter and React Native for cross-platform developers
• Any GGUF Model: Supports any GGUF model from Huggingface (Qwen, Gemma, Llama, DeepSeek, etc.)
• Multi-Modal AI: Run LLMs, VLMs, Embedding Models, TTS models and more
• Optimized Performance: From FP32 to as low as 2-bit quantized models for efficiency
• Agentic: More performant workflows with mobile tool calling
• Native Support: iOS xcframework and JNILibs for native setup
• Tiny C++ Build: For custom hardware deployments
• Advanced Features: Chat templates with Jinja2 support and token streaming

Quick Start

Choose your preferred platform:

Platform Examples

Note: due to divergent framework support, the initialization patterns for Flutter and React Native are slightly different:

• React Native CactusLM is initialized with a local model file (inside the app sandbox)
• Flutter CactusLM is initialized with a HuggingFace download URL

import 'package:cactus/cactus.dart';

final lm = await CactusLM.init(
  modelUrl: 'https://huggingface.co/Cactus-Compute/Qwen3-600m-Instruct-GGUF/resolve/main/Qwen3-0.6B-Q8_0.gguf',
  contextSize: 2048,
);

final messages = [ChatMessage(role: 'user', content: 'Hello!')];
final response = await lm.completion(messages, maxTokens: 100, temperature: 0.7);

import { CactusLM } from 'cactus-react-native';
import RNFS from 'react-native-fs'; // install RNFS for file management

const filePath = `${RNFS.DocumentDirectoryPath}/${fileName}`;

const { lm, error } = await CactusLM.init({
  model: filePath,
  n_ctx: 2048,
});

const messages = [{ role: 'user', content: 'Hello!' }];
const params = { n_predict: 100, temperature: 0.7 };
const response = await lm.completion(messages, params);

common_params params;
params.model.path = 'path/to/your/model.gguf';
context.loadModel(params)

context.params.prompt = "Hello, how are you?";
context.params.n_predict = 100;
context.initSampling()

context.beginCompletion();
context.loadPrompt();

while (context.has_next_token && !context.is_interrupted) {
  auto token_output = context.doCompletion();
  if (token_output.tok == -1) break;
}

Get started by watching a quickstart video and building one of our example apps:

• Flutter
• React Native

Telemetry

Cactus offers powerful telemetry for all your React Native projects.

To take advantage of your Cactus telemetry, visit our React Native documentation .

Performance Benchmarks

Real-world performance on popular mobile devices:

Demo Apps

Try our demo applications to see Cactus SDK in action:

Next Steps

Community

• Join our Discord - Get help and connect with other developers
• Visualize Repository - Explore the codebase structure
• GitHub Repository - View source code and contribute