Olmo-3.1-32B-Instruct

48 of 64 layers cache only a 4,096-token window rather than the full context, on a period of . Figures assume the default configuration; --swa-full disables the saving entirely.

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 16.89 GiB. The real file is 18.14 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 8.00 GiB at 32K context where the real figure is 2.84 GiB, because most of this model's layers cache a fixed window rather than the whole context.

How much VRAM does Olmo-3.1-32B-Instruct need?

Q4_K_M is exactly 19,482,034,976 bytes (18.14 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.

How large is Olmo-3.1-32B-Instruct's KV cache?

2.84 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 Olmo-3.1-32B-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.