Yes — Phi-5 Medium 14B (14B) runs at 48 tok/s on M4 Pro with 24 GB RAM using Q4_K_M quantization via MLX. First token latency is 0.7s. A capable open-source LLM with 14B parameters.
LLMCheck measured Phi-5 Medium 14B on M4 Pro using the standard methodology: Q4_K_M quantization, 256-token input, 512-token output, 3 runs averaged on a freshly-booted system.
| Metric | Value |
|---|---|
| Tokens per second | 48 tok/s |
| Time to first token | 0.7s |
| Quantization | Q4_K_M |
| Minimum RAM | 24 GB |
| Recommended engine | MLX |
| Parameters | 14B |
| Benchmark date | 2026-06 |
Q4_K_M 14B MLX M4 Pro
The recommended engine for Phi-5 Medium 14B on M4 Pro is MLX. Install with pip and pull the model:
Alternatively, you can use Ollama for a simpler setup:
| Chip | Speed | First Token | Min RAM | Engine |
|---|---|---|---|---|
| M5 Max | 65 tok/s | 0.5s | 128 GB | MLX |
| M5 Max | 58 tok/s | 0.6s | 64 GB | Ollama |
| M3 | 32 tok/s | 1.1s | 16 GB | Ollama |
| M2 | 24 tok/s | 1.4s | 16 GB | Ollama |
To run Phi-5 Medium 14B on M4 Pro you need:
See how Phi-5 Medium 14B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard