456 GB/s (+256)
B
Raises estimated decode speed by about 57%.55.1 tok/s decode
Raises estimated decode speed by about 57%.
~$249 MSRP
Can Mistral 7B Instruct v0.3 run on MacBook Pro M2 Pro 16GB?
YES — Runs Great
B66Good
○Estimated from fit model
Mistral 7B Instruct v0.3 needs ~8.9 GB VRAM. MacBook Pro M2 Pro 16GB has 11.5 GB. With Q4_K_M quantization, expect ~35 tok/s.
Runtime: llama.cppCapacity: RoomyBandwidth: Very lowStack: StandardBottleneck: Memory bandwidth
Share:
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.
Select quantization to explore
Q4_K_M (Medium quality) — 8.9 GB, 35.2 tok/s, Runs well
8.9 GB required11.5 GB available
77% VRAM used
Fit status
Runs well
Decode
35.2 tok/s
TTFT
5493 ms
Safe context
8K
Memory
8.9 GB / 11.5 GB
Memory breakdown
Weights4.3 GB
KV Cache2.0 GB
Runtime0.9 GB
Headroom1.7 GB
See how fast it feels
See how fast it feelsMistral 7B Instruct v0.3 on MacBook Pro M2 Pro 16GB
1st promptCold start — includes initialization
>Solve: If a train travels at 60 mph for 2.5 hours, how far does it go?
Let me work through this step by step.
Given information:
- Speed = 60 mph (miles per hour)
- Time = 2.5 hours
Formula: Distance = Speed x Time
Calculation:
Distance = 60 mph x 2.5 h
= 150 miles
The train travels 150 miles.
Double-check: at 60 mph the train covers 60 miles each hour. In half an hour it covers 30 miles. So in 2.5 hours: 60 + 60 + 30 = 150 miles. Confirmed.
2nd promptWarm — model ready, no init overhead
>What if the train stops for 20 minutes halfway?
The stop doesn't change the total distance covered while moving, but it affects total trip time.
Breakdown:
- First half: 1.25 hours at 60 mph = 75 miles
- Stop: 20 minutes (no distance)
- Second half: 1.25 hours at 60 mph = 75 miles
Total distance: 150 miles (same)
Total time: 2.5 h + 0.33 h = 2.83 hours
Average speed: 150 / 2.83 = 53 mph
The distance stays the same but average speed drops to 53 mph because of the stop.
Estimated: 35.2 tok/s decode · 5.5s TTFT (warm) · 88 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 | B | Runs well | 35.2 tok/s | 2996 ms | 8K |
| Coding | B | Runs well | 35.2 tok/s | 5493 ms | 8K |
| Agentic Coding | B | Tight fit | 35.2 tok/s | 7990 ms | 8K |
| Reasoning | B | Runs well | 35.2 tok/s | 6492 ms | 8K |
| RAG | B | Tight fit | 35.2 tok/s | 9987 ms | 8K |
Quantization options
How Mistral 7B Instruct v0.3 (7B params) fits at each quantization level on MacBook Pro M2 Pro 16GB (11.5 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 2.7 GB | Low | B62 |
Q3_K_S | 3 | 3.4 GB | Low | B62 |
NVFP4 | 4 |
Get started
Copy-paste commands to run Mistral 7B Instruct v0.3 on your machine.
Run
lms load Mistral-7B-Instruct-v0.3 && lms server startUpgrade options
