How much VRAM does LongWriter-Zero-32B need?

Its published quantizations range from 9.27 GiB to 61.04 GiB in weights alone, before KV cache and runtime overhead.

Can I run LongWriter-Zero-32B on a GeForce RTX 5050?

Q: Can I run LongWriter-Zero-32B on a GeForce RTX 5050?

No. No published quantization of LongWriter-Zero-32B fits a GeForce RTX 5050 at any context with q4_0 KV cache.

Not at these settings. No indexed quantization of LongWriter-Zero-32B fits GeForce RTX 5050 at any context we compute, with q4_0 KV. The smallest shipped quantization is 9.27 GiB in weights alone, against 7.44 GiB usable. CPU offload can still run it, slowly.

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◐
BF16	61.04 GiB	62.2	62.5	63.1	64.2	66.4	70.9
Q8_0	32.43 GiB	33.6	33.9	34.5	35.6	37.8	42.3
Q6_K_L	25.39 GiB	26.6	26.9	27.4	28.5	30.8	35.3
Q6_K	25.04 GiB	26.2	26.5	27.1	28.2	30.4	34.9
Q5_K_L	22.11 GiB	23.3	23.6	24.1	25.3	27.5	32.0
Q5_K_M	21.66 GiB	22.8	23.1	23.7	24.8	27.1	31.6
Q5_K_S	21.08 GiB	22.3	22.5	23.1	24.2	26.5	31.0
Q4_1	19.22 GiB	20.4	20.7	21.2	22.4	24.6	29.1
Q4_K_L	19.03 GiB	20.2	20.5	21.1	22.2	24.4	28.9
Q4_K_M	18.49 GiB	19.7	19.9	20.5	21.6	23.9	28.4
Q4_K_S	17.49 GiB	18.7	19.0	19.5	20.6	22.9	27.4
Q4_0	17.43 GiB	18.6	18.9	19.5	20.6	22.8	27.3
IQ4_NL	17.40 GiB	18.6	18.9	19.4	20.5	22.8	27.3
IQ4_XS	16.64 GiB	17.8	18.1	18.7	19.8	22.0	26.5
Q3_K_L	16.06 GiB	17.2	17.5	18.1	19.2	21.5	26.0
Q3_K_M	14.84 GiB	16.0	16.3	16.9	18.0	20.2	24.7
IQ3_M	13.79 GiB	15.0	15.3	15.8	16.9	19.2	23.7
Q3_K_S	13.40 GiB	14.6	14.9	15.4	16.6	18.8	23.3
IQ3_XS	12.76 GiB	13.9	14.2	14.8	15.9	18.2	22.7
Q2_K_L	12.18 GiB	13.4	13.6	14.2	15.3	17.6	22.1
IQ3_XXS	11.96 GiB	13.1	13.4	14.0	15.1	17.4	21.9
Q2_K	11.47 GiB	12.6	12.9	13.5	14.6	16.9	21.4
IQ2_M	10.49 GiB	11.7	12.0	12.5	13.6	15.9	20.4
IQ2_S	9.67 GiB	10.9	11.1	11.7	12.8	15.1	19.6
IQ2_XS	9.27 GiB	10.5	10.7	11.3	12.4	14.7	19.2

●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.

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, most of this model's layers cache only a 131,072-token window rather than the full context.