🚀 Echo-1 (0.5B Parameters - GGUF)

Echo-1 is an edge-optimized 0.5B parameter model based on the robust Qwen-2.5-Instruct architecture. Fine-tuned via LoRA and fully merged into a single, high-performance GGUF binary, this model balances lightning-fast local inference speeds with advanced structural reasoning and context tracking.

✨ Key Features

• Qwen-2.5 Foundation: Inherits deep tokenization efficiency, vastly improved instruction following, and stable textual coherence from the base architecture.
• Merged Standalone Build: Zero dependencies on external adapter weights, multi-layer configs, or Python runtimes—loadable instantly in any standard GGUF engine (llama.cpp, node-llama-cpp, Ollama).
• Ultra-Low Memory Footprint: Extremely lightweight structure makes it ideal for local-first computing environments, private automated utilities, and background processes on low-spec consumer systems.

---

🧠 Prompt Structure (ChatML Syntax)

Because Echo-1 is fine-tuned on top of Qwen-2.5-Instruct, it utilizes standard ChatML template markers. For precise structural alignment and to prevent response cutoff, construct your inputs exactly like this:

<|im_start|>system
You are Echo-1, a helpful assistant.<|im_end|>
<|im_start|>user
Write a short paragraph explaining the benefits of local-first AI.<|im_end|>
<|im_start|>assistant

---

💻 Sample Implementation (Node.js)

You can spin this model up locally using node-llama-cpp. Ensure you append the raw ChatML structural sequences directly to your execution queries:

import {LlamaModel, LlamaContext, LlamaSequence} from "node-llama-cpp";
import path from "path";

const model = new LlamaModel({
    modelPath: path.join(__dirname, "echo-1-0.5b.gguf")
});

const context = new LlamaContext({model});
const sequence = new LlamaSequence({context});

const prompt = `<|im_start|>system\nYou are Echo-1.<|im_end|>\n<|im_start|>user\nWhat is 15 + 27?<|im_end|>\n<|im_start|>assistant\n`;
const tokens = model.tokenize(prompt);

console.log("Echo-1 response:");
const response = await sequence.evaluate(tokens);
console.log(model.detokenize(response));

---

📄 License

This model's merged weights are distributed under the Apache 2.0 License, strictly adhering to the foundational terms and commercial/private usage permissions granted by the original Qwen team.