How much VRAM does Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness need?

Its published quantizations range from 4.88 GiB to 5.09 GiB in weights alone, before KV cache and runtime overhead.

Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness on GeForce RTX 3050

Can I run Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness on a GeForce RTX 3050?

Q: Can I run Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness on a GeForce RTX 3050?

No. No published quantization of Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness fits a GeForce RTX 3050 at any context with f16 KV cache.

Not at these settings. No indexed quantization of Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness fits GeForce RTX 3050 at any context we compute, with f16 KV. The smallest shipped quantization is 4.88 GiB in weights alone, against 5.58 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— 6 GB card, 5.58 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◐
Q4_K_M	5.09 GiB	6.0	6.1	6.2	6.5	7.0	8.0
Q4_K_S	4.88 GiB	5.8	5.9	6.0	6.2	6.7	7.7

●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, the KV cache grows linearly with context and, past about 32K, becomes larger than the weights for many models.