How much VRAM does command-a-plus-05-2026-bf16 need?

Its published quantizations range from 45.87 GiB to 216.03 GiB in weights alone, before KV cache and runtime overhead.

command-a-plus-05-2026-bf16 on Radeon RX 6500 XT

Can I run command-a-plus-05-2026-bf16 on a Radeon RX 6500 XT?

Q: Can I run command-a-plus-05-2026-bf16 on a Radeon RX 6500 XT?

No. No published quantization of command-a-plus-05-2026-bf16 fits a Radeon RX 6500 XT at any context with q8_0 KV cache.

Not at these settings. No indexed quantization of command-a-plus-05-2026-bf16 fits Radeon RX 6500 XT at any context we compute, with q8_0 KV. The smallest shipped quantization is 45.87 GiB in weights alone, against 3.72 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— 4 GB card, 3.72 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◐
Q8_0	216.03 GiB	217.2	217.3	217.4	217.7	218.2	219.3
Q6_K	176.79 GiB	178.0	178.1	178.2	178.5	179.0	180.0
Q5_K_M	146.93 GiB	148.1	148.2	148.3	148.6	149.1	150.2
Q5_K_S	141.69 GiB	142.9	143.0	143.1	143.3	143.9	144.9
Q4_1	129.29 GiB	130.5	130.6	130.7	131.0	131.5	132.5
Q4_K_L	126.06 GiB	127.2	127.3	127.5	127.7	128.2	129.3
Q4_K_M	125.81 GiB	127.0	127.1	127.2	127.5	128.0	129.1
Q4_K_S	120.92 GiB	122.1	122.2	122.3	122.6	123.1	124.2
Q4_0	117.42 GiB	118.6	118.7	118.8	119.1	119.6	120.7
IQ4_NL	116.79 GiB	118.0	118.1	118.2	118.5	119.0	120.0
IQ4_XS	110.67 GiB	111.8	111.9	112.1	112.3	112.9	113.9
IQ3_M	99.41 GiB	100.6	100.7	100.8	101.1	101.6	102.7
Q3_K_L	99.27 GiB	100.4	100.5	100.7	100.9	101.5	102.5
Q3_K_M	95.53 GiB	96.7	96.8	96.9	97.2	97.7	98.8
IQ3_XS	95.16 GiB	96.3	96.4	96.6	96.8	97.3	98.4
Q3_K_S	91.28 GiB	92.5	92.5	92.7	92.9	93.5	94.5
IQ3_XXS	87.48 GiB	88.6	88.7	88.9	89.1	89.7	90.7
Q2_K_L	74.80 GiB	76.0	76.1	76.2	76.5	77.0	78.1
Q2_K	74.56 GiB	75.7	75.8	76.0	76.2	76.7	77.8
IQ2_M	71.32 GiB	72.5	72.6	72.7	73.0	73.5	74.6
IQ2_S	65.10 GiB	66.3	66.4	66.5	66.8	67.3	68.4
IQ2_XS	63.98 GiB	65.1	65.2	65.4	65.6	66.2	67.2
IQ2_XXS	57.98 GiB	59.1	59.2	59.4	59.6	60.2	61.2
IQ1_M	50.59 GiB	51.8	51.9	52.0	52.2	52.8	53.8
IQ1_S	45.87 GiB	47.0	47.1	47.3	47.5	48.1	49.1

●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 4,096-token window rather than the full context.