Mistral Medium 3.5 (128B) on Apple M3 Ultra (96 GB)

Can the Apple M3 Ultra (96 GB) run Mistral Medium 3.5 (128B)?

Yes, it runs well

Runs well. At Q2_K it needs about 60.4 GB of 72.0 GB usable, with an estimated speed of 11.1 tok/s.

Best quant: Q2_K
Generation speed: 11.1 tok/s
Prompt processing: 138 tok/s
First token (1K prompt): 7.4 s
Largest safe context: 39K

Best app to run it on the Apple M3 Ultra (96 GB)

On macOS

LM Studio

Desktop app · runs on llama.cpp, MLX · use the MLX build

Runs the MLX build, typically up to 10% faster than GGUF for a model with 127.7B active parameters on Apple silicon.
Point-and-click: search, download and chat in one window.

Open Mistral Medium 3.5 (128B) in LM Studio

Ollama One command with Ollama’s own build of this model.
llama.cpp Usually the first to support new model architectures.

Want an agent? Mistral Medium 3.5 (128B) supports tool use, so you can pair a llama.cpp or MLX LM server with one of these. Apps that run on this computer come first.

Coding agents

Personal assistants

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download Mistral Medium 3.5 (128B)

Recommended for the Apple M3 Ultra (96 GB): Q2_K. Every other quantization has its own download link in the table below.

Quickest · Ollama

One command downloads a ready-to-run version and starts a chat.

Get Ollama at ollama.com/download (opens in a new tab) · model page (opens in a new tab)

ollama run mistral-medium-3.5:128b

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the Q2_K file (opens in a new tab)

huggingface.co/unsloth/Mistral-Medium-3.5-128B-GGUF (opens in a new tab)

ollama run hf.co/unsloth/Mistral-Medium-3.5-128B-GGUF:Q2_K

Original weights · Hugging Face

Full-precision files from Mistral AI, for fine-tuning, vLLM, Transformers.

huggingface.co/mistralai/Mistral-Medium-3.5-128B (opens in a new tab)

No sign-in needed.

Official repo · GitHub

Reference code, inference scripts and docs from Mistral AI. The weights themselves download from Hugging Face.

github.com/mistralai/mistral-inference (opens in a new tab)

Mistral AI’s official inference code.

git clone https://github.com/mistralai/mistral-inference

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q2_K, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Runs well: How well it runs on the Apple M3 Ultra (96 GB): speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	25.3 / 45
Memory fitFull marks when the model fits in the share of unified memory the GPU can use; fewer when it goes past that limit.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	7.9 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	15.0 / 15
Match score	73.2 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the Apple M3 Ultra (96 GB)

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	60.4 GB	72.0 GB	11.1 tok/s	Runs well	56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q3_K_M`	71.9 GB	72.0 GB	8.8 tok/s	Slowtight fit	67Good quality (B) · 67 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q4_K_M`	87.2 GB	72.0 GB	5.8 tok/s	Slowfar over GPU memory limit	76Great quality (A) · 76 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 59/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q5_K_M`	100.0 GB	72.0 GB	—	Won't run	77Great quality (A) · 77 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 10/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q6_K`	114 GB	72.0 GB	—	Won't run	79Great quality (A) · 79 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 10/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q8_0`	145 GB	72.0 GB	—	Won't run	80Great quality (A) · 80 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 10/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`F16`	265 GB	72.0 GB	—	Won't run	80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 10/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab) Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

76Great quality (A) · 76 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA RTX PRO 6000 Blackwell Q4_K_M · 17.2 tok/s · Runs well (tight fit)
56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. Apple M2 Ultra (192 GB) Q2_K · 10.9 tok/s · Runs well
56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. Apple M3 Ultra (256 GB) Q2_K · 11.1 tok/s · Runs well
56Good quality (B) · 56 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking. Apple M3 Ultra (512 GB) Q2_K · 11.1 tok/s · Runs well
80Great quality (A) · 80 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA B200 Q8_0 · 39.1 tok/s · Runs great
77Great quality (A) · 77 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA H200 141GB SXM Q5_K_M · 38.1 tok/s · Runs great

Quick answers

How much memory does Mistral Medium 3.5 (128B) need on the M3 Ultra 96GB?

At Q2_K it needs about 60.4 GB. The M3 Ultra 96GB has 72.0 GB usable.

How fast is Mistral Medium 3.5 (128B) on the M3 Ultra 96GB?

About 11.1 tok/s at Q2_K, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Mistral Medium 3.5 (128B) should I use on the M3 Ultra 96GB?

Q2_K. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.