Yes — Phi-5 Large 28B (28B) runs at 38 tok/s on M5 Max with 128 GB RAM using Q4_K_M quantization via MLX. First token latency is 0.6s. A capable open-source LLM with 28B parameters.
LLMCheck measured Phi-5 Large 28B on M5 Max 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 | 38 tok/s |
| Time to first token | 0.6s |
| Quantization | Q4_K_M |
| Minimum RAM | 128 GB |
| Recommended engine | MLX |
| Parameters | 28B |
| Benchmark date | 2026-07 |
Q4_K_M 28B MLX M5 Max
The recommended engine for Phi-5 Large 28B on M5 Max 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 |
|---|---|---|---|---|
| M4 Pro | 28 tok/s | 0.9s | 32 GB | MLX |
| M3 Max | 26 tok/s | 1.1s | 64 GB | Ollama |
To run Phi-5 Large 28B on M5 Max you need:
Phi-5 Large 28B needs about 128 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
Not sure which Mac fits your budget? See the best Mac for running this →
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep our benchmarks free.
See how Phi-5 Large 28B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard