Microsoft releases Phi-4.5: 14B parameter SLM with best-in-class reasoning, runs on 8GB VRAM
In one sentence Microsoft releases Phi-4.5, a 14-billion parameter model that outperforms much larger models on reasoning and coding benchmarks, runs on a laptop GPU with 8GB VRAM, and is freely available under Apache 2.0.
Imagine hiring a world-class chess grandmaster who can fit in your backpack instead of needing an entire warehouse to operate. That is essentially what Microsoft has done with Phi-4.5.
AI models usually follow a simple rule: bigger means smarter. Phi-4.5 breaks that rule in a surprising way. With just 14 billion parameters, it competes with models that are two to four times its size on logical reasoning and code-writing tasks.
The genuinely exciting part for everyday computer users is that Phi-4.5 runs on a graphics card with just 8GB of video memory. That is the kind of GPU found in many mid-range gaming laptops and desktops. No expensive server required, no internet connection needed — the model runs entirely on your own machine.
The 128,000-token context window means the model can read and reason over very long documents in a single session, like an entire codebase or a lengthy research paper.
Microsoft chose the Apache 2.0 license, the most permissive in open source: anyone can use it, modify it, and even embed it in commercial products at no cost.
Installing it is as simple as running a single command with Ollama, a tool that makes downloading an AI model about as easy as installing an app. More technical users can leverage the ONNX format to plug it into existing Windows applications or data pipelines.
In short: Microsoft has built a powerful model that can finally walk through the front door of your home office without needing a forklift.
Companies
Microsoft
Tools
Phi-4.5, Ollama, ONNX
Tags
Sources