Mac Studio M3 Ultra 96GBBudget pick
96 GB Unified (+48)
C
Adds memory headroom for longer context windows and future model growth.112 tok/s decode
Adds memory headroom for longer context windows and future model growth.
~$3,999 MSRP
Can Dolphin3.0 Llama3.1 8B run on Radeon Pro W7900 48GB?
YES — Runs Great
C47Usable
○Estimated from fit model
Dolphin3.0 Llama3.1 8B needs ~11.5 GB VRAM. Radeon Pro W7900 48GB has 48.0 GB. With Q4_K_M quantization, expect ~105 tok/s.
Runtime: llama.cppCapacity: RoomyBandwidth: HighStack: StandardBottleneck: Balanced
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) — 11.5 GB, 104.5 tok/s, Runs well
11.5 GB required48.0 GB available
24% VRAM used
Fit status
Runs well
Decode
104.5 tok/s
TTFT
1853 ms
Safe context
639K
Memory
11.5 GB / 48.0 GB
Memory breakdown
Weights4.9 GB
KV Cache0.9 GB
Runtime0.9 GB
Headroom4.8 GB
See how fast it feels
See how fast it feelsDolphin3.0 Llama3.1 8B on Radeon Pro W7900 48GB
1st promptCold start — includes initialization
>What is local AI inference?
Local AI inference means running an AI language model directly on your own hardware — your laptop, desktop, or server — instead of sending requests to a remote cloud API.
When you run inference locally the model weights are loaded into your GPU or unified memory. Each token you generate requires reading those weights from memory, so memory bandwidth is the main bottleneck for decode speed.
Key benefits of running locally:
- Full privacy: your prompts never leave your machine
- No per-token cost or rate limits
- Works offline once the model is downloaded
- Latency depends only on your hardware
2nd promptWarm — model ready, no init overhead
>How much VRAM do I need?
It depends on the model size and quantization level. A rough rule of thumb:
Model size Q4 (4-bit) Q8 (8-bit) FP16
7B params ~4.3 GB ~7.5 GB ~14 GB
13B params ~7.9 GB ~13.9 GB ~26 GB
70B params ~42.7 GB ~74.9 GB ~140 GB
Most people use 4-bit quantization (Q4_K_M) which gives 90-95% of full quality at a fraction of the memory. A 24 GB GPU can comfortably run most 7B-13B models.
Estimated: 104.5 tok/s decode · 1.9s TTFT (warm) · 261 tok/s prefill
What limits this setup
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
Best improvement path
Performance by workload
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | C | Runs well | 104.5 tok/s | 1011 ms | 639K |
| Coding | C | Runs well | 104.5 tok/s | 1853 ms | 639K |
| Agentic Coding | C | Runs well | 104.5 tok/s | 2696 ms | 639K |
| Reasoning | C | Runs well | 104.5 tok/s | 2190 ms | 639K |
| RAG | C | Runs well | 104.5 tok/s | 3370 ms | 639K |
Quantization options
How Dolphin3.0 Llama3.1 8B (8B params) fits at each quantization level on Radeon Pro W7900 48GB (48.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 3.1 GB | Low | C42 |
Q3_K_S | 3 | 3.9 GB | Low | C42 |
NVFP4 | 4 |
Get started
Copy-paste commands to run Dolphin3.0 Llama3.1 8B on your machine.
Run
lms load hf-dphn--dolphin3-0-llama3-1-8b-gguf && lms server startUpgrade options