Can I run Qwen3-1.7B-Distilled-30B-A3B-SFT on a 8GB card?

Yes. Qwen3-1.7B-Distilled-30B-A3B-SFT runs on a 8GB card: the largest quantization that fits is F16 at 16,384 context with f16 KV, needing 6.33 GiB of 7.44 GiB usable memory.

How much VRAM does Qwen3-1.7B-Distilled-30B-A3B-SFT need?

Its published quantizations range from 1.19 GiB to 3.79 GiB in weights alone, before KV cache and runtime overhead.

Qwen3-1.7B-Distilled-30B-A3B-SFT on 8GB card

Does Qwen3-1.7B-Distilled-30B-A3B-SFT fit in 8GB of VRAM?

Yes. The best fit is F16 at 16,384 context with f16 KV — 6.33 GiB of 7.44 GiB usable, leaving 1.11 GiB headroom.

From the file· weights summed from filesFrom the file· KV computed per layerPredicted· speed and compute buffer

KV cache dtypef16 q8_0 q4_0— 8 GB card, 7.44 GiB usable after overhead

Every quantization at every context

total memory required; green fits, red does not

Quant	Weights●	4K◐	8K◐	16K◐	32K◐	64K◐	128K◐
F16	3.79 GiB	5.0	5.5	6.3	8.1	11.6	18.6
Q8_0	2.02 GiB	3.2	3.7	4.6	6.3	9.8	16.8
Q5_K_M	1.37 GiB	2.6	3.0	3.9	5.7	9.2	16.2
Q4_K_M	1.19 GiB	2.4	2.9	3.7	5.5	9.0	16.0

●From the file◐Predictedwhat these mean

Figures are GiB of total memory: weights plus KV cache plus compute buffer and backend overhead. Weights and KV are near-exact; the overhead term is modeled. Hover any cell for the breakdown.

Run it

The best-fitting configuration above, as a command:

llama-cli -hf reaperdoesntknow/Qwen3-1.7B-Distilled-30B-A3B-SFT \
  --ctx-size 16384 \
  -ngl auto

Recent llama.cpp defaults to --fit on with -ngl auto, so it will size the offload for you. The question worth your attention is not how many layers to offload but what context and quantization you are willing to live with — which is what the grid above is for.

Why there are no speeds on this page

A capacity is not a card. Whether a model fits depends only on memory, so every figure above holds for any 8GB accelerator. How fast it runs depends on memory bandwidth, which varies several-fold between cards of the same capacity — so putting a tokens-per-second number here would be inventing one. Pick a specific card from hardware and the speed column appears.

Why other calculators disagree

A parameters × bits ÷ 8 estimate ignores two things that dominate at long context. First, the weights themselves are not the nominal rate — quantizations are mixtures, so the real file is consistently larger than the label implies. Second, the KV cache grows linearly with context and, past about 32K, becomes larger than the weights for many models.