NVIDIA H200 141GB经济之选141 GB VRAM (+13)4800 GB/s (+4527)
C
Raises estimated decode speed by about 1073%.210 tok/s 解码
Raises estimated decode speed by about 1073%.
Adds memory headroom for longer context windows and future model growth.
~$30,000 MSRP
Can Phi 4 reasoning vision 15B run on NVIDIA DGX Spark 128GB?
YES — With F16
C43Usable
Estimated from fit model
Phi 4 reasoning vision 15B needs ~46.8 GB VRAM. NVIDIA DGX Spark 128GB has 0 MB. With F16 quantization, expect ~8 tok/s.
Runtime: OllamaCapacity: RoomyBandwidth: LowStack: BasicBottleneck: Memory bandwidth
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.
Phi 4 reasoning vision 15B at Q4_K_M needs 12.1 GB — too much for NVIDIA DGX Spark 128GB (0.0 GB). Runs at F16 (46.8 GB) with maximum quality. 8 quantization levels fit.
Select quantization to explore
Q4_K_M (Medium quality) — 25.2 GB, 17.9 tok/s, Runs well
25.2 GB required108.8 GB available
23% VRAM used
Fit status
Runs well
Decode
17.9 tok/s
TTFT
10815 ms
Safe context
777K
Memory
25.2 GB / 108.8 GB
Memory breakdown
Weights9.2 GB
KV Cache1.8 GB
Runtime1.2 GB
Headroom13.1 GB
See how fast it feels
See how fast it feelsPhi 4 reasoning vision 15B on NVIDIA DGX Spark 128GB
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: 17.9 tok/s decode · 10.8s TTFT (warm) · 45 tok/s prefill
What limits this setup
The model fits in shared memory, but shared-memory bandwidth is now the real limiter.
Fit does not mean dedicated-VRAM speed
Unified or shared memory can make a model technically fit, but sustained tokens per second may still trail a discrete high-bandwidth GPU with less total memory.
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
Prioritize bandwidth, not only capacity
If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.
Performance by workload
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | C | Runs well | 17.9 tok/s | 5899 ms | 777K |
| Coding | F | Too heavy | 3.2 tok/s | 60081 ms | 4K |
| Agentic Coding | C | Runs well | 17.9 tok/s | 15730 ms | 777K |
| Reasoning | C | Runs well | 17.9 tok/s | 12781 ms | 777K |
| RAG | C | Runs well | 17.9 tok/s | 19663 ms | 777K |
Quantization options
How Phi 4 reasoning vision 15B (15B params) fits at each quantization level on NVIDIA DGX Spark 128GB (92.2 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 5.9 GB | Low | D39 |
Q3_K_S | 3 | 7.4 GB | Low | D39 |
NVFP4 | 4 | 8.4 GB | Medium | D39 |
Q4_K_M | 4 | 9.2 GB | Medium | D39 |
Q5_K_M | 5 | 10.8 GB | High | D40 |
Q6_K | 6 | 12.3 GB | High | D40 |
Q8_0 | 8 | 16.1 GB | Very High | C40 |
F16Best for your GPU | 16 | 30.7 GB | Maximum | C42 |
Get started
Copy-paste commands to run Phi 4 reasoning vision 15B on your machine.
Run
lms load hf-jamesburton--phi-4-reasoning-vision-15b-gguf && lms server start升级选项
能流畅运行 Phi 4 reasoning vision 15B 的硬件
NVIDIA H200 141GB经济之选NVIDIA H200 PCIe 141GB最佳性价比141 GB VRAM (+13)4800 GB/s (+4527)
C
Raises estimated decode speed by about 1073%.210 tok/s 解码
Raises estimated decode speed by about 1073%.
Adds memory headroom for longer context windows and future model growth.
~$30,000 MSRP
NVIDIA B200 180GBNVIDIA 升级180 GB VRAM (+52)8000 GB/s (+7727)
C
Raises estimated decode speed by about 1073%.210 tok/s 解码
Raises estimated decode speed by about 1073%.
Adds memory headroom for longer context windows and future model growth.
~$30,000 MSRP
Frequently asked questions
Can NVIDIA DGX Spark 128GB run Phi 4 reasoning vision 15B?
Yes, NVIDIA DGX Spark 128GB can run Phi 4 reasoning vision 15B at F16 quantization (Runs well). The recommended Q4_K_M requires 12.1 GB which exceeds available memory, but at F16 it needs only 46.8 GB. Expected decode speed: 7.5 tok/s.
How much VRAM does Phi 4 reasoning vision 15B need?
Phi 4 reasoning vision 15B (15B parameters) requires approximately 12.1 GB at Q4_K_M quantization. On NVIDIA DGX Spark 128GB, it fits at F16 using 46.8 GB.
What is the best quantization for Phi 4 reasoning vision 15B?
The recommended quantization is Q4_K_M, but on NVIDIA DGX Spark 128GB the best fitting quantization is F16, which uses 46.8 GB.
What speed will Phi 4 reasoning vision 15B run at on NVIDIA DGX Spark 128GB?
On NVIDIA DGX Spark 128GB, Phi 4 reasoning vision 15B achieves approximately 7.5 tokens per second decode speed with a time-to-first-token of 25960ms using F16 quantization.
Can NVIDIA DGX Spark 128GB run Phi 4 reasoning vision 15B for coding?
For coding workloads, Phi 4 reasoning vision 15B on NVIDIA DGX Spark 128GB receives a F grade with 3.2 tok/s and 4K context.
What context window can Phi 4 reasoning vision 15B use on NVIDIA DGX Spark 128GB?
On NVIDIA DGX Spark 128GB, Phi 4 reasoning vision 15B can safely use up to 581K tokens of context at F16 quantization. The model's official context limit is —, but available memory constrains the safe maximum.
What should I upgrade first if Phi 4 reasoning vision 15B feels slow on NVIDIA DGX Spark 128GB?
Prioritize bandwidth, not only capacity. If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.
Is unified memory on NVIDIA DGX Spark 128GB as fast as VRAM for Phi 4 reasoning vision 15B?
Not always. NVIDIA DGX Spark 128GB can often fit larger models thanks to unified memory, but a discrete GPU with dedicated high-bandwidth VRAM may still decode faster once the model fits. For this combination, the important distinction is capacity versus sustained throughput.
Embed this result▼
Paste this snippet into any page to show a live fit card.
<iframe src="https://willitrunai.com/embed/hf-jamesburton--phi-4-reasoning-vision-15b-gguf-on-dgx-spark-128gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>
Preview: