What GPU do I need to run NousResearch/Hermes-3-Llama-3.1-405B-FP8?
A 406B language model for chat and instruction-following. 405.9B parameters, published in F8_E4M3. View on Hugging Face
Hermes-3-Llama-3.1-405B-FP8 is published by NousResearch on Hugging Face, with 150 downloads and 29 likes to date. It's a LlamaForCausalLM model built for text-generation, published natively in F8_E4M3.
What Hermes-3-Llama-3.1-405B-FP8 is
Hermes-3-Llama-3.1-405B-FP8 is a 406B-parameter language model published by Nous Research on Hugging Face. It is released under llama3.
License note: check the model's own license terms on Hugging Face before commercial use. Facts in this section are sourced from Hermes-3-Llama-3.1-405B-FP8's Hugging Face model card, not benchmarked by Aquanode.
What it's used for
- Chat assistants
- Instruction following
- Synthetic data generation
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| FP8 (native) | 378.0 GB | 453.6 GB | AMD MI300X | 3 | $7.89/hr |
| INT4 (quantized) | 189.0 GB | 226.8 GB | RTX A6000 | 5 | $1.81/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Hermes-3-Llama-3.1-405B-FP8 at its published (F8_E4M3) precision: 3× AMD MI300X, at $2.63/hr per GPU ($7.89/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Hermes-3-Llama-3.1-405B-FP8: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Grouped-query attention: every layer caches keys and values for a few shared KV heads. Cached per token: 516,096 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 2.0 GB |
| 32K | 15.8 GB |
| 128K | 63.0 GB |
Computed from the layer, head and window counts in the model's own configuration (NousResearch/Hermes-3-Llama-3.1-405B-FP8), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
Hermes-3-Llama-3.1-405B-FP8: common questions
Can Hermes-3-Llama-3.1-405B-FP8 run on a single GPU?
No. At FP8 (native) it needs 453.6 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 192.0 GB AMD MI300X, and it takes 3 of them.
Is Hermes-3-Llama-3.1-405B-FP8 already quantized?
Yes. It is published in FP8, one byte per parameter, so the 453.6 GB figure is already a quantized footprint rather than a full-precision one. Only INT4 goes below it, at 226.8 GB. A GPU without FP8 tensor cores cannot run it as published, which is why cards here are matched on precision support and not on VRAM alone.
How many GPUs do I need to run Hermes-3-Llama-3.1-405B-FP8?
3 at FP8 (native). It needs 453.6 GB of VRAM and the cheapest capable live offer is a 192.0 GB AMD MI300X, so 3 of them come to $7.89/hr in total.
Does quantizing Hermes-3-Llama-3.1-405B-FP8 lower the GPU bill?
Yes. At FP8 (native) the cheapest live fit is 3 AMD MI300X cards at $7.89/hr. At INT4 (quantized) it drops to 5 RTX A6000 cards at $1.81/hr, provided a quantized checkpoint exists for it.
How to run Hermes-3-Llama-3.1-405B-FP8
Run Hermes-3-Llama-3.1-405B-FP8 with vLLM
Generic example, not from the model's own docs: adjust flags (quantization, context length, parallelism) for your setup.
vllm serve NousResearch/Hermes-3-Llama-3.1-405B-FP8 --tensor-parallel-size 3Deploy Hermes-3-Llama-3.1-405B-FP8 on Aquanode
Aquanode has no one-click deploy template for Hermes-3-Llama-3.1-405B-FP8; you install the inference engine yourself with the commands below. Aquanode sells GPU pods billed per second, not a hosted inference API.
- Launch a bare GPU pod sized to the requirement above (3× AMD MI300X or larger).
- Open a terminal on the pod, or save one of the commands above as a startup script so it runs automatically the first time the pod boots.
- Run the command and connect to the resulting endpoint.
Launch a GPU podWeight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More Hermes 3 models
- Hermes-3-Llama-3.1-70B (70.6B, BF16)
- Hermes-3-Llama-3.1-8B (8.0B, BF16)
More on Hermes-3-Llama-3.1-405B-FP8
Fits on a 288 GB GPU at INT4: every model that fits in 288 GB.
Related reading: AMD MI300X pricing and specs, How much VRAM you need for LLMs, and What AI inference is.
