Back to search results
Ministral-3-8B-Instruct-2512 · IQ4_XS
Apple M5 Max128GBOllama v0.23L2 Cross-framework verified
This page aggregates 1 real-world runs of Ministral-3-8B-Instruct-2512 (IQ4_XS) on Apple M5 Max with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.
67.21 tok/s
Decode
Decode speed
271.49 tok/s
Prefill
Prefill speed
1.92 s
TTFT
Time to first token
14.53 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
12 days ago
Last verified
Performance
- llama.cpp · IQ4_XS Decode 75.66 · Prefill 303.23 ·
- Ollama · IQ4_XS (current)Decode 67.21 · Prefill 271.49 ·
Core figures
- Decode (avg)
- 67.21 tok/s
- Prefill (avg)
- 271.49 tok/s
- TTFT (avg)
- 1.92 s
- VRAM (avg)
- 14.53 GB
- MTP acceptance rate
- —
- TTFB
- 0.54 GB
- Power draw
- 46.5 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Ministral-3-8B-Instruct-2512
- Quantization
- IQ4_XS
- Framework
- Ollama
- Version
- v0.23
- Context length
- 65536 tokens
- Batch size
- 256
- GPU layers
- 20
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M5 Max
- Nominal VRAM
- 128 GB
- Measured VRAM (avg)
- 14.53 GB
- OS
- macOS 26.1
- Power draw
- 46.5 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.23 · macOS 26.1 · 65536 ctx
67.21 tok/s
Decode
271.49 tok/s
Prefill
1.92 s
TTFT
14.53 GB
VRAM
—
MTP
46.5 W
Power draw
L2 batch seed 2026-10-24: Ollama on hw_id=28 running model_id=74 at IQ4_XS