Silicon Score Apple Silicon local LLM rankings
← Canonical rankings

Silicon Score › Chips › M4 Ultra (192 GB)

M4 Ultra (192 GB) — benchmark record

M4 Ultra (192 GB) local LLM benchmarks on Apple Silicon Mac with 8 published rows across 8 models. Peak published speed is 45.0 tok/s on Mistral Small 4 119B. Published runtimes include MLX, Ollama.

8Benchmark rows
8Models tested
45.0Fastest avg tok/s (Mistral Small 4 119B)
0Lab benchmarks

Quick take

Best published speed here is 45.0 tok/s on Mistral Small 4 119B at Q4_K - Medium. This page is evidence, not the full buying answer.

Based on 8 external benchmarks; no lab runs yet.

Published runtimes: MLX, Ollama.

Need a recommendation? Use BuyChecking a target model? Use FitWant a setup answer? Use RunBrowse Macs by exact hardwareNeed the full audit trail? Use Bench

Current published coverage

Published model coverage includes Mistral Small 4 119B, Qwen 3 32B, Llama 4 Scout 17B-16E, Qwen 3 235B-A22B plus 4 more published models. Published runtime coverage on this chip includes MLX, Ollama. Fastest published row is 45.0 tok/s on Mistral Small 4 119B at Q4_K - Medium.

Mistral Small 4 119BQwen 3 32BLlama 4 Scout 17B-16EQwen 3 235B-A22BLlama 3.3 70BDeepSeek R1 Distill Llama 70B

Raw benchmark rows for M4 Ultra (192 GB)

Rows sorted by avg tok/s descending. Click source badge to see original measurement page.

ModelQuantRAM req.ContextAvg tok/sPrompt tok/sRuntimeSource
Mistral Small 4 119BQ4_K - Medium——45.0 tok/s—MLXref
Qwen 3 32BQ4_K - Medium——32.0 tok/s—Ollamaref
Llama 4 Scout 17B-16EQ4_K - Medium——30.0 tok/s—MLXref
Qwen 3 235B-A22BQ4_K - Medium——22.0 tok/s—MLXref
Llama 3.3 70BQ4_K - Medium——18.0 tok/s—MLXref
DeepSeek R1 Distill Llama 70BQ4_K - Medium——16.0 tok/s—MLXref
Qwen 2.5 72BQ4_K - Medium——15.0 tok/s—Ollamaref
gpt-oss 120BQ4_K - Medium——10.0 tok/s—MLXref

Models tested on this chip

Mistral Small 4 119BQwen 3 32BLlama 4 Scout 17B-16EQwen 3 235B-A22BLlama 3.3 70BDeepSeek R1 Distill Llama 70BQwen 2.5 72Bgpt-oss 120B

Next step

Use Rankings when you need the best overall answer for a Mac, and keep this page open when you need the evidence behind that recommendation.

If you are comparing local Apple Silicon against rented cloud hardware, use AI Data Center Index for current GPU rental context.

Data

benchmarks.json — full dataset  ·  chips.json — chip summaries  ·  benchmarks.csv — CSV export

Data from in-house lab measurements plus community-published benchmarks. See all chip families →