What Models?

Pick the right model for your GPU — in seconds.

Phi-4 Mini 3.8B VRAM requirements: which GPUs can run it locally?

3.84B params 128K max context

The smallest GPU memory that runs Phi-4 Mini 3.8B well is 6 GB (Q8_0, e.g. GeForce RTX 3050 6GB). 128 of 132 GPUs run it well and 3 more are a tight fit.

MMLU 70

VRAM needed per quantization

QuantizationWeights16K context32K contextFull 128K
Q4_K_M2.3 GB4.3 GB6.2 GB17.9 GB
Q8_03.8 GB5.8 GB7.7 GB19.4 GB

Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.

Run it locally

Each command downloads the GGUF from Hugging Face, then Ollama starts a chat and llama.cpp starts an OpenAI-compatible server with a web UI on port 8080.

Ollama ollama run hf.co/unsloth/Phi-4-mini-instruct-GGUF:Q4_K_M
llama.cpp llama-server -hf unsloth/Phi-4-mini-instruct-GGUF:Q4_K_M
Ollama ollama run hf.co/unsloth/Phi-4-mini-instruct-GGUF:Q8_0
llama.cpp llama-server -hf unsloth/Phi-4-mini-instruct-GGUF:Q8_0

Cheapest GPUs that can run Phi-4 Mini 3.8B

Lowest launch MSRP among GPUs that run its smallest quantization (Q4_K_M) well, with at least 16K context.

As an Amazon Associate, I earn from qualifying purchases.

GPUs that can run Phi-4 Mini 3.8B

Best quantization that fits on each GPU, with the context and speed it gets. For chips with several memory sizes, the smallest that works is shown.

NVIDIA

AMD

Intel

Apple

  • M1 8 GB Q8_0 34K context · ~13 tok/s
  • M2 8 GB Q8_0 34K context · ~20 tok/s
  • M3 8 GB Q8_0 34K context · ~20 tok/s
  • M1 Pro 16 GB Q8_0 100K context · ~39 tok/s
  • M2 Pro 16 GB Q8_0 100K context · ~39 tok/s
  • M4 16 GB Q8_0 100K context · ~24 tok/s
  • M5 16 GB Q8_0 100K context · ~30 tok/s
  • M6 16 GB Q8_0 100K context · ~30 tok/s
  • M3 Pro 18 GB Q8_0 116K context · ~30 tok/s
  • M4 Pro 24 GB Q8_0 128K context · ~54 tok/s
  • M5 Pro 24 GB Q8_0 128K context · ~60 tok/s
  • M1 Max 32 GB Q8_0 128K context · ~79 tok/s
  • M2 Max 32 GB Q8_0 128K context · ~79 tok/s
  • M3 Max 36 GB Q8_0 128K context · ~59 tok/s
  • M4 Max 36 GB Q8_0 128K context · ~81 tok/s
  • M5 Max 36 GB Q8_0 128K context · ~91 tok/s
  • M1 Ultra 64 GB Q8_0 128K context · ~110 tok/s
  • M2 Ultra 64 GB Q8_0 128K context · ~110 tok/s
  • M3 Ultra 96 GB Q8_0 128K context · ~113 tok/s
  • M5 Ultra 96 GB Q8_0 128K context · ~165 tok/s
Check your exact setup in the full tool →