How much VRAM does Huihui-gpt-oss-20b-BF16-abliterated need?

Its published quantizations range from 10.70 GiB to 88.57 GiB in weights alone, before KV cache and runtime overhead.

Huihui-gpt-oss-20b-BF16-abliterated on GeForce RTX 2080 Ti

Can I run Huihui-gpt-oss-20b-BF16-abliterated on a GeForce RTX 2080 Ti?

Q: Can I run Huihui-gpt-oss-20b-BF16-abliterated on a GeForce RTX 2080 Ti?

No. No published quantization of Huihui-gpt-oss-20b-BF16-abliterated fits a GeForce RTX 2080 Ti at any context with q4_0 KV cache.

Not at these settings. No indexed quantization of Huihui-gpt-oss-20b-BF16-abliterated fits GeForce RTX 2080 Ti at any context we compute, with q4_0 KV. The smallest shipped quantization is 10.70 GiB in weights alone, against 10.23 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— 11 GB card, 10.23 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◐
Q5_1	88.57 GiB	89.4	89.4	89.5	89.6	89.8	90.2
IQ4_NL	78.58 GiB	79.4	79.4	79.5	79.6	79.8	80.2
F16	38.99 GiB	39.8	39.8	39.9	40.0	40.2	40.6
BF16	38.99 GiB	39.8	39.8	39.9	40.0	40.2	40.6
Q8_0	20.73 GiB	21.6	21.6	21.6	21.7	21.9	22.4
Q4_K_M	14.72 GiB	15.5	15.6	15.6	15.7	15.9	16.4
Q3_K_M	12.03 GiB	12.8	12.9	12.9	13.0	13.2	13.7
Q6_K	11.21 GiB	12.0	12.1	12.1	12.2	12.4	12.9
Q6_K_L	11.21 GiB	12.0	12.1	12.1	12.2	12.4	12.9
Q5_K_L	11.09 GiB	11.9	11.9	12.0	12.1	12.3	12.7
Q4_K_L	11.07 GiB	11.9	11.9	12.0	12.1	12.3	12.7
Q2_K_L	11.03 GiB	11.9	11.9	11.9	12.0	12.3	12.7
Q5_K_M	10.92 GiB	11.7	11.8	11.8	11.9	12.1	12.6
Q5_K_S	10.92 GiB	11.7	11.8	11.8	11.9	12.1	12.6
Q4_K_S	10.87 GiB	11.7	11.7	11.8	11.9	12.1	12.5
Q4_1	10.80 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ4_XS	10.77 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ3_M	10.77 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ3_XS	10.77 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ3_XXS	10.77 GiB	11.6	11.6	11.7	11.8	12.0	12.4
Q2_K	10.77 GiB	11.6	11.6	11.7	11.8	12.0	12.4
Q3_K_S	10.76 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ2_M	10.75 GiB	11.6	11.6	11.7	11.8	12.0	12.4
IQ2_S	10.75 GiB	11.6	11.6	11.7	11.8	12.0	12.4
Q4_0	10.73 GiB	11.5	11.6	11.6	11.7	11.9	12.4
IQ2_XS	10.72 GiB	11.5	11.6	11.6	11.7	11.9	12.4
IQ2_XXS	10.72 GiB	11.5	11.6	11.6	11.7	11.9	12.4
Q3_K_L	10.70 GiB	11.5	11.5	11.6	11.7	11.9	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 128-token window rather than the full context.