Can Codestral RAG 19B Pruned i1 run on Mac Studio M2 Ultra 128GB?

YES — Runs Great

C45Usable

○Estimated from fit model

Codestral RAG 19B Pruned i1 needs ~28.5 GB VRAM. Mac Studio M2 Ultra 128GB has 92.2 GB. With Q4_K_M quantization, expect ~40 tok/s.

Runtime: llama.cppCapacity: RoomyBandwidth: HighStack: StandardBottleneck: Balanced

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 28.5 GB, 40.0 tok/s, Runs well

28.5 GB required92.2 GB available

31% VRAM used

Fit status

Runs well

Decode

40.0 tok/s

TTFT

4836 ms

Safe context

473K

Memory

28.5 GB / 92.2 GB

Memory breakdown

Weights11.6 GB

KV Cache2.2 GB

Runtime0.9 GB

Headroom13.8 GB

See how fast it feels

See how fast it feelsCodestral RAG 19B Pruned i1 on Mac Studio M2 Ultra 128GB
1st promptCold start — includes initialization
>Write a Python function to sort a list
Here's a Python function that sorts a list using quicksort:

def quicksort(arr):
 if len(arr) <= 1:
 return arr
 pivot = arr[len(arr) // 2]
 left = [x for x in arr if x < pivot]
 middle = [x for x in arr if x == pivot]
 right = [x for x in arr if x > pivot]
 return quicksort(left) + middle + quicksort(right)

# Example usage
numbers = [3, 6, 8, 10, 1, 2, 1]
print(quicksort(numbers)) # [1, 1, 2, 3, 6, 8, 10]

The algorithm picks a pivot, partitions the list into smaller, equal, and larger elements, then recursively sorts each partition and concatenates the results.
2nd promptWarm — model ready, no init overhead
>Now add type hints and a docstring
def quicksort(arr: list[int]) -> list[int]:
 """Sort a list using the quicksort algorithm.

 Args:
 arr: List of integers to sort.

 Returns:
 A new sorted list.
 """
 if len(arr) <= 1:
 return arr
 pivot = arr[len(arr) // 2]
 left = [x for x in arr if x < pivot]
 middle = [x for x in arr if x == pivot]
 right = [x for x in arr if x > pivot]
 return quicksort(left) + middle + quicksort(right)

Added type hints for the input and return type, plus a Google-style docstring.
Estimated: 40.0 tok/s decode · 4.8s TTFT (warm) · 100 tok/s prefill

What limits this setup

This setup is broadly balanced for this model.

Shared-memory contention still exists

The OS, browser, and inference runtime all compete for the same physical memory pool, so real-world headroom is less forgiving than raw capacity suggests.

Best improvement path

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	C	Runs well	40.0 tok/s	2638 ms	473K
Coding	C	Runs well	40.0 tok/s	4836 ms	473K
Agentic Coding	C	Runs well	40.0 tok/s	7034 ms	473K
Reasoning	C	Runs well	40.0 tok/s	5715 ms	473K
RAG	C	Runs well	40.0 tok/s	8792 ms	473K

Quantization options

How Codestral RAG 19B Pruned i1 (19B params) fits at each quantization level on Mac Studio M2 Ultra 128GB (92.2 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	7.4 GB	Low	D39
Q3_K_S	3	9.3 GB	Low	D39
NVFP4	4	10.6 GB	Medium	D39
Q4_K_M	4	11.6 GB	Medium	D39
Q5_K_M	5	13.7 GB	High	D39
Q6_K	6	15.6 GB	High	D40
Q8_0	8	20.3 GB	Very High	C40
F16Best for your GPU	16	38.9 GB	Maximum	C44

Get started

Copy-paste commands to run Codestral RAG 19B Pruned i1 on your machine.

Run

lms load hf-mradermacher--codestral-rag-19b-pruned-i1-gguf && lms server start

Upgrade options

Hardware that runs Codestral RAG 19B Pruned i1 well

👁 NVIDIA
RTX PRO 6000 Blackwell Workstation Edition 96GBBudget pick

1792 GB/s (+992)

Raises estimated decode speed by about 225%.129.9 tok/s decode

Raises estimated decode speed by about 225%.

~$9,999 MSRP

👁 NVIDIA
RTX PRO 6000 Blackwell Server Edition 96GBBest value

1597 GB/s (+797)

Raises estimated decode speed by about 189%.115.7 tok/s decode

Raises estimated decode speed by about 189%.