- Vendor
- LiquidAI
- Quantization
- MLX 4-bit
- Parameters
- 1.2B
- Size
- 632.8 MB
$ brew install mirai$ mirai --model LiquidAI/LFM2.5-1.2B-Instruct-MLX-4bitBenchmarks
LFM 2.5
Apple M4 Max 128GB
0.22 s
lower is better ↓
uzu0.5.14vs
0.31.2vs
llama.cpp0.3.23LFM 2.5
Apple M4 Max 128GB
530 t/s
higher is better ↑
uzu0.5.14vs0.31.2vsllama.cpp0.3.23LFM 2.5
Apple M4 Max 128GB
0.79 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
LFM2.5-1.2B-Instruct-MLX-4bit is a compact, quantized version of Liquid AI's LFM2.5-1.2B-Instruct model, exported specifically for fast inference on Apple Silicon devices using the MLX framework. At just 628 MB on disk, it delivers a remarkably efficient edge-deployment option for multilingual text generation.
Architecture & Specifications
This is a 4-bit quantized export (group size 64) of the 1.2-billion-parameter LFM2.5 instruction-tuned model. Despite its small footprint, it supports a generous 128K token context length, making it well-suited for tasks that require processing or generating long-form content on local hardware.
Multilingual Capability
LFM2.5-1.2B supports ten languages out of the box: English, Japanese, Korean, French, Spanish, German, Italian, Portuguese, Arabic, and Chinese — providing broad multilingual coverage for an edge-class model.
Ideal Use Cases
- On-device chat and instruction following on MacBooks, iMacs, and other Apple Silicon machines
- Low-latency local inference where cloud connectivity is unavailable or undesirable
- Multilingual text generation for lightweight assistants and embedded applications
- Long-context workloads that benefit from the 128K context window without requiring a large GPU
Getting Started
The model integrates directly with the `mlx-lm` Python library. Liquid AI recommends conservative sampling parameters — low temperature (0.1), top-k of 50, top-p of 0.1, and a slight repetition penalty of 1.05 — to produce focused, high-quality outputs.
Provenance
Developed by Liquid AI and derived from the base LFM2.5-1.2B-Instruct model. Released under the LFM 1.0 License.