Yes — Hermes 4 70B (70B) runs at 18 tok/s on M4 Ultra with 192 GB RAM using Q4_K_M quantization via Ollama. First token latency is 1.5s. A capable open-source LLM with 70B parameters.
LLMCheck measured Hermes 4 70B on M4 Ultra 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 | 18 tok/s |
| Time to first token | 1.5s |
| Quantization | Q4_K_M |
| Minimum RAM | 192 GB |
| Recommended engine | Ollama |
| Parameters | 70B |
| Benchmark date | 2026-05 |
Q4_K_M 70B Ollama M4 Ultra
The recommended engine for Hermes 4 70B on M4 Ultra is Ollama. Install Ollama, then pull the model:
Ollama handles quantization automatically — it will download the Q4_K_M variant (~192 GB) and start an interactive chat session.
| Chip | Speed | First Token | Min RAM | Engine |
|---|---|---|---|---|
| M5 Max | 16 tok/s | 1.8s | 128 GB | MLX |
| M4 Max | 13 tok/s | 2.1s | 128 GB | MLX |
To run Hermes 4 70B on M4 Ultra you need:
See how Hermes 4 70B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard