- Vendor
- LiquidAI
- Quantization
- MLX 8-bit
- Parameters
- 2.6B
- Size
- 2.5 GB
$ brew install mirai$ mirai --model mlx-community/LFM2-2.6B-8bitBenchmarks
LFM 2
Apple M4 Max 128GB
0.51 s
lower is better ↓
uzu0.5.14vs
0.31.2vs
llama.cpp0.3.23LFM 2
Apple M4 Max 128GB
156 t/s
higher is better ↑
uzu0.5.14vs0.31.2vsllama.cpp0.3.23LFM 2
Apple M4 Max 128GB
2.75 GB
lower is better ↓
uzu0.5.14vs0.31.2vsllama.cpp0.3.23Benchmarked 7 Aug 2026
Integrate with SDK
https://github.com/trymirai/uzu-swift
| 1 | import Foundation |
| 2 | import Uzu |
| 3 | |
| 4 | public func runChat() async throws { |
| 5 | let engineConfig = EngineConfig.create() |
| 6 | let engine = try await Engine.create(config: engineConfig) |
| 7 | |
| 8 | guard let model = try await engine.model(identifier: "alibaba:qwen3.5:0.8b:mirai:mirai-m:4") else { |
| 9 | return |
| 10 | } |
| 11 | for try await update in try await engine.download(model: model).iterator() { |
| 12 | print(String(format: "\r\u{001B}[2KDownload progress: %.2f%%", update.progress() * 100), terminator: "") |
| 13 | fflush(stdout) |
| 14 | } |
| 15 | print() |
| 16 | |
| 17 | let messages = [ |
| 18 | ChatMessage.system().withText(text: "You are a helpful assistant"), |
| 19 | ChatMessage.user().withText(text: "Tell me a short, funny story about a robot") |
| 20 | ] |
| 21 | let session = try await engine.chat(model: model, config: .create()) |
| 22 | let stream = await session.replyWithStream(input: messages, config: .create()) |
| 23 | var message: ChatMessage? = nil |
| 24 | for try await update in stream.iterator() { |
| 25 | switch update { |
| 26 | case .replies(let replies): |
| 27 | let reply = replies.last |
| 28 | message = reply?.message |
| 29 | print("Generated tokens: \(reply?.stats.tokensCountOutput ?? 0)") |
| 30 | case .error(let error): |
| 31 | print("Error: \(error)") |
| 32 | } |
| 33 | } |
| 34 | print("Reasoning: \(message?.reasoning() ?? "empty")") |
| 35 | print("Text: \(message?.text() ?? "empty")") |
| 36 | } |
| 37 | |
Details
An 8-bit quantized version of LiquidAI's LFM2-2.6B, converted to Apple's MLX format for efficient on-device inference on Apple Silicon hardware. This community conversion enables fast, memory-friendly text generation directly on Mac devices.
Origin & Architecture
LFM2-2.6B is developed by Liquid AI as part of their LFM2 (Liquid Foundation Model) series, designed specifically for edge deployment. The model sits at 2.6 billion parameters — compact enough for local use while still delivering capable language generation. This variant was quantized to 8-bit precision and converted to the MLX framework using `mlx-lm` v0.28.0.
Multilingual Support
The model supports eight languages out of the box:
- English, French, German, Spanish
- Arabic, Chinese, Japanese, Korean
This broad language coverage makes it a versatile choice for multilingual applications running on-device without cloud dependencies.
Key Strengths
- Edge-optimized: Purpose-built for constrained environments, making it ideal for local deployment on laptops and desktops powered by Apple Silicon.
- 8-bit quantization: Reduces memory footprint while preserving model quality, allowing smooth inference even on machines with limited unified memory.
- MLX-native: Takes full advantage of Apple's MLX framework for optimized performance on M-series chips.
Use Cases
LFM2-2.6B-8bit is well suited for local text generation tasks including conversational AI, content drafting, multilingual translation assistance, and lightweight coding support — all without requiring a network connection or external API. Its small footprint and chat-template support make it a practical choice for developers building private, responsive AI-powered applications on macOS.
License: LFM 1.0 (custom license from Liquid AI)