Computed

Hermes 4.3 36B on an A100 80GB (SXM)

Yes, at full precision, though context is limited to 25k tokens.

Change the numbers →Hermes 4.3 36B on every card A100 80GB (SXM) across every model

Verdict

Fits at FP16

73.6 GiB usable of 80 GB

Weights (Q4_K_M)

20.3 GiB

67.3 GiB at FP16 · 36.2B params · 1 of 6 quant rows are measured files

KV cache

256.0 KiB per token at FP16

64 layers × 8 KV heads × 128 dimensions

Speed ceiling

27 tok/s

2039 GB/s ÷ bytes read per token

Every quant, against this card

Weight bytes are the size of the real published file wherever one exists — 1 of these6 rows are measured from the published checkpoint, the rest computed from the parameter count. Max context is what the KV cache can grow to in whatever memory the weights leave behind, capped at the 512k tokens this model was trained to address.

Bars are the weight bytes at each precision; the dashed line is the usable memory of an A100 80GB (SXM). Drawn from the computed byte counts, not sketched.

Precision	Bits/weight	Weights	Source	Fits	Max context	Ceiling
FP16 / BF16	16.00	67.3 GiB	measured file	yes	25,654	27 tok/s
Q8_0	8.50	35.8 GiB	nominal	yes	154,940	50 tok/s
Q6_K	6.56	27.6 GiB	nominal	yes	188,383	64 tok/s
Q5_K_M	5.67	23.9 GiB	nominal	yes	203,725	73 tok/s
Q4_K_M	4.83	20.3 GiB	nominal	yes	218,205	85 tok/s
Q3_K_M	3.91	16.5 GiB	nominal	yes	234,064	103 tok/s

What the context actually costs

The KV cache is 2 · layers · kv_heads · head_dim bytes per token per element — 2 · 64 · 8 · 128 · 2 B = 256.0 KiB. It depends on the key/value head count, not the attention-head count and not the parameter count. This model shares 8 KV heads across 80 query heads, which divides the cache by 10 against multi-head attention.

Context	KV cache (FP16)	KV cache (8-bit)	Plus Q4_K_M weights
4,096	1.0 GiB	512 MiB	21.3 GiB
8,192	2.0 GiB	1.0 GiB	22.3 GiB
32,768	8.0 GiB	4.0 GiB	28.3 GiB
131,072	32.0 GiB	16.0 GiB	52.3 GiB

The speed ceiling, and where it comes from

Generating one token reads every active weight from memory once. At FP16 / BF16 that is 67.3 GiB, plus a pass over the KV cache. An A100 80GB (SXM) moves 2039 GB/s, so the arithmetic ceiling is 27 tok/s. Treat it as a bound, not an estimate: attention overhead, kernel launches and imperfect memory access keep real runtimes at roughly 60–80% of it, and nothing pushes past it.

Where these numbers come from

The model

Parameters: 36,151,104,512
Layers: 64
Attention / KV heads: 80 / 8
Head dimension: 128
Trained context: 524,288
Checkpoint as published: 67.3 GiB

Read from NousResearch/Hermes-4.3-36B. The parameter count is the Hub's own total over the tensor shapes, not a figure taken from the model's name.

The accelerator

Memory: 80 GB HBM2e
Bandwidth: 2039 GB/s
Assumed usable: 92% → 73.6 GiB

Capacity and bandwidth from the vendor's specification. The usable fraction is an assumption, not a spec: a driver context, compute workspace and any attached display come out of the same pool before a weight is loaded.

Questions this pairing answers

How much VRAM does Hermes 4.3 36B need?

67.3 GiB for the weights at FP16 — 36.2B parameters at two bytes each — and 20.3 GiB at Q4_K_M. The KV cache is on top of that and is not a fixed number: this model spends 256.0 KiB per token of context, so 8,192 tokens costs a further 2.0 GiB. An A100 80GB (SXM) makes 73.6 GiB of its 80 GB available on the assumption below.

Can an A100 80GB (SXM) run Hermes 4.3 36B?

Yes, at full precision, though context is limited to 25k tokens. That is the weights and the KV cache together, against 73.6 GiB of usable memory.

How fast will Hermes 4.3 36B run on an A100 80GB (SXM)?

No faster than 27 tokens/second at FP16 / BF16, and in practice below it. Decoding is memory-bound: every token reads all 67.3 GiB of weights plus the cache, and this card moves 2039 GB/s. That division is the ceiling — no kernel, runtime or driver beats it, and a real runtime typically reaches 60–80% of it.

Why does the context length change how much memory Hermes 4.3 36B needs?

Because the KV cache holds one key and one value vector per token, per layer, for the whole conversation, and it is allocated separately from the weights. This model has 64 layers and 8 key/value heads of 128 dimensions, shared across 80 query heads — grouped-query attention, which divides the cache by 10. That works out at 256.0 KiB per token. Parameter count tells you nothing about this number.

The same model on a different card

A different model on the same card

All 62 models on A100 80GB (SXM) →

Method and limits. Weight bytes are the byte size of the real published file wherever one exists, and the model's exact parameter count times the published llama.cpp bits-per-weight where it does not. That distinction is on every row above and it matters at both ends: a sub-1B model's Q4_K_M file runs a third larger than the nominal figure because k-quants keep its embedding tables at higher precision, and an already-4-bit release cannot be quantized upward at all. The KV cache is 2 · layers · kv_heads · head_dim · bytes per token, summed over layers with each sliding-window layer capped at its window. The speed figure is a roofline bound, not a benchmark: bandwidth divided by bytes read per token, which no runtime exceeds and every runtime falls short of. The usable fraction of card memory is an assumption: 92% here, which is what this lane assumes for a dedicated card — the other class assumes 75%, so it is not one number applied to every device. That is the only assumed input on this page; every other figure is computed from the model config and the card's published specification. Nothing on this page is written by a language model. Architecture from the model's published config, fetched 2026-08-07.