Vikhr-Gemma-2B-instruct

13 of 26 layers cache only a 4,096-token window rather than the full context, on a period of 2. Figures assume the default configuration; --swa-full disables the saving entirely.

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 1.37 GiB. The real file is 1.59 GiB, because a quantization is a mixture and some tensors are always kept at higher precision. The larger discrepancy is the cache: a flat formula gives 3.25 GiB at 32K context where the real figure is 1.85 GiB, because most of this model's layers cache a fixed window rather than the whole context.

How much VRAM does Vikhr-Gemma-2B-instruct need?

Q4_K_M is exactly 1,708,582,848 bytes (1.59 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.

How large is Vikhr-Gemma-2B-instruct's KV cache?

1.85 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.

Which quantization of Vikhr-Gemma-2B-instruct should I use?

Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.