llama-3.3-70b-Instruct-4bit on M3 Ultra (60-core GPU) + 96GB unified
MeasuredOn 2026-04-28, llama-3.3-70b-Instruct-4bit ran at 16.8 tok/s (chat-short) on an M3 Ultra (60-core GPU) with mlx. Run r_sx3a4y9n-m4.
M3 Ultra (60-core GPU) + 96GB unified
suite suite-v1
cli 0.0.1-dev
signedG8xb3zMu3+…
Embed badgesubmitted Apr 28, 2026Workload results
| Workload | Backend | Model | decode tok/s | prefill tok/s | TTFT | p50 | p95 |
|---|---|---|---|---|---|---|---|
| chat-short | mlx@0.31.3 | mlx-community/llama-3.3-70b-Instruct-4bit | 16.78tok/s | 25.09tok/s | 5,420ms | 59.6ms | 59.8ms |
Reproduce on your machine
Start with this model and workload selection. This command does not capture the original artifact, backend version or server settings:
Match quantization, context capacity, caching and runtime settings for a useful comparison. Duration varies by workload and hardware; the published client may differ from this run's version. How it's measured.
Share this run
Copy the short link, a Reddit-ready summary, or a README badge. Every number is the measured result on this page, and each badge links back to it.
https://llm-speed.com/r/r_sx3a4y9n-m4Related benchmarks
- More llama-3.3-70b-Instruct-4bit benchmarks — every backend and rig that has run this model.
- More M3 Ultra (60-core GPU) LLM benchmarks — every model measured on this hardware.
Provenance
- Run ID
- r_sx3a4y9n-m4
- Fingerprint hash
- bbf15132ccbbe7d7
- Public key
- G8xb3zMu3+pznEici/TiW0gPk5qSNIYIikGCwm1rMdQ=
- Received
- 2026-04-28 14:45:36