How much VRAM does Laguna-XS-2.1 need?

Its published quantizations range from 8.76 GiB to 62.33 GiB in weights alone, before KV cache and runtime overhead.

Can I run Laguna-XS-2.1 on a GeForce RTX 4060 Ti?

Q: Can I run Laguna-XS-2.1 on a GeForce RTX 4060 Ti?

No. No published quantization of Laguna-XS-2.1 fits a GeForce RTX 4060 Ti at any context with q8_0 KV cache.

Not at these settings. No indexed quantization of Laguna-XS-2.1 fits GeForce RTX 4060 Ti at any context we compute, with q8_0 KV. The smallest shipped quantization is 8.76 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	62.33 GiB	63.3	63.4	63.5	63.9	64.5	65.9
Q8_0	33.15 GiB	34.1	34.2	34.3	34.7	35.3	36.7
Q6_K_L	27.14 GiB	28.1	28.2	28.3	28.7	29.3	30.7
Q6_K	27.04 GiB	28.0	28.1	28.2	28.6	29.2	30.6
Q5_K_L	22.47 GiB	23.4	23.5	23.7	24.0	24.7	26.0
Q5_K_M	22.36 GiB	23.3	23.4	23.6	23.9	24.5	25.9
Q5_K_S	21.63 GiB	22.6	22.7	22.8	23.2	23.8	25.1
Q4_1	19.73 GiB	20.7	20.8	20.9	21.3	21.9	23.2
Q4_K_L	19.28 GiB	20.2	20.3	20.5	20.8	21.5	22.8
Q4_K_M	19.14 GiB	20.1	20.2	20.3	20.7	21.3	22.7
Q4_K_S	18.42 GiB	19.4	19.5	19.6	19.9	20.6	21.9
IQ4_NL	17.89 GiB	18.8	18.9	19.1	19.4	20.1	21.4
Q4_0	17.89 GiB	18.8	18.9	19.1	19.4	20.1	21.4
IQ4_XS	16.96 GiB	17.9	18.0	18.2	18.5	19.2	20.5
IQ3_M	15.16 GiB	16.1	16.2	16.4	16.7	17.4	18.7
Q3_K_L	15.02 GiB	16.0	16.1	16.2	16.5	17.2	18.5
Q3_K_M	14.51 GiB	15.5	15.5	15.7	16.0	16.7	18.0
IQ3_XS	14.49 GiB	15.4	15.5	15.7	16.0	16.7	18.0
Q3_K_S	13.87 GiB	14.8	14.9	15.1	15.4	16.1	17.4
IQ3_XXS	13.30 GiB	14.2	14.3	14.5	14.8	15.5	16.8
Q2_K_L	11.51 GiB	12.5	12.5	12.7	13.0	13.7	15.0
Q2_K	11.32 GiB	12.3	12.4	12.5	12.8	13.5	14.8
IQ2_M	10.84 GiB	11.8	11.9	12.0	12.4	13.0	14.4
IQ2_S	9.89 GiB	10.8	10.9	11.1	11.4	12.1	13.4
IQ2_XS	9.67 GiB	10.6	10.7	10.9	11.2	11.9	13.2
IQ2_XXS	8.76 GiB	9.7	9.8	10.0	10.3	10.9	12.3

●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 512-token window rather than the full context.