What GPU do I need to run deepseek-ai/DeepSeek-V3?
A 685B (MoE) language model for chat and instruction-following. 684.5B parameters, published in F8_E4M3. View on Hugging Face
DeepSeek-V3 is published by deepseek-ai on Hugging Face, with 1,078,839 downloads and 4,177 likes to date. It's a DeepseekV3ForCausalLM model built for text-generation, published natively in F8_E4M3.
What DeepSeek-V3 is
DeepSeek-V3 is a 685B-parameter mixture-of-experts language model published by DeepSeek on Hugging Face. It reads images alongside text. It is released under Not stated.
License note: no license tag published on the model's Hugging Face card; check the repo directly before any commercial use. Facts in this section are sourced from DeepSeek-V3'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) | 637.5 GB | 765.0 GB | AMD MI300X | 4 | $10.52/hr |
| INT4 (quantized) | 318.8 GB | 382.5 GB | RTX A6000 | 8 | $2.90/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 DeepSeek-V3 at its published (F8_E4M3) precision: 4× AMD MI300X, at $2.63/hr per GPU ($10.52/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
DeepSeek-V3: 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.
Multi-head latent attention: every layer caches one compressed latent per token. Cached per token: 70,272 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.27 GB |
| 32K | 2.1 GB |
| 128K | 8.6 GB |
| 160K (model maximum) | 10.7 GB |
Computed from the layer, head and window counts in the model's own configuration (deepseek-ai/DeepSeek-V3), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
DeepSeek-V3: common questions
Can DeepSeek-V3 run on a single GPU?
No. At FP8 (native) it needs 765.0 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 4 of them.
Is DeepSeek-V3 already quantized?
Yes. It is published in FP8, one byte per parameter, so the 765.0 GB figure is already a quantized footprint rather than a full-precision one. Only INT4 goes below it, at 382.5 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 DeepSeek-V3?
4 at FP8 (native). It needs 765.0 GB of VRAM and the cheapest capable live offer is a 192.0 GB AMD MI300X, so 4 of them come to $10.52/hr in total.
Does quantizing DeepSeek-V3 lower the GPU bill?
Yes. At FP8 (native) the cheapest live fit is 4 AMD MI300X cards at $10.52/hr. At INT4 (quantized) it drops to 8 RTX A6000 cards at $2.90/hr, provided a quantized checkpoint exists for it.
How to run DeepSeek-V3
Run DeepSeek-V3 with vLLM
Generic example, not from the model's own docs: adjust flags (quantization, context length, parallelism) for your setup.
vllm serve deepseek-ai/DeepSeek-V3 --tensor-parallel-size 4Run DeepSeek-V3 with GGUF quantizations
Prebuilt GGUF weights published at bullerwins/DeepSeek-V3-GGUF. Run with llama.cpp's llama-server or load the repo directly in LM Studio.
llama-server -hf bullerwins/DeepSeek-V3-GGUFDeploy DeepSeek-V3 on Aquanode
Aquanode has no one-click deploy template for DeepSeek-V3; 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 (4× 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 DeepSeek V3 models
- DeepSeek-V3-0324 (684.5B, F8_E4M3)
- DeepSeek-V3.1 (684.5B, F8_E4M3)
- DeepSeek-V3.1-Base (684.5B, F8_E4M3)
- DeepSeek-V3.1-Terminus (684.5B, F8_E4M3)
- DeepSeek-V3.2 (685.4B, F8_E4M3)
Lineage
Also published as: unsloth/DeepSeek-V3. Identical copies of this model under another account.
Alternatives at this size
Other models for text-generation within about a third of DeepSeek-V3's 684.5B parameters, from other model lines.
- GLM-5.2 (753.3B, BF16)
- NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 (560.5B, BF16)
- Kimi-K2-Instruct (1026.4B, F8_E4M3)
- A.X-K2 (691.7B, F8_E4M3)
- MiMo-V2.5-Pro (1023.2B, F8_E4M3)
More on DeepSeek-V3
Best chat and assistants models: how DeepSeek-V3 ranks against the rest.
Compare DeepSeek-V3 with GLM-5.2, NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16, Kimi-K2-Instruct, Llama-3.1-405B.
Related reading: AMD MI300X pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
