LLM GPU requirements

Llama 3.1 8B GPU requirements: VRAM, GPUs and servers

How much GPU memory Llama 3.1 8B needs in BF16, FP8 and 4-bit, how its KV cache grows with context, what fine-tuning takes, and which servers run it, with monthly prices.

Parameters: 8B
Architecture: Dense 32 layers
KV heads: 8 × 128 dim
Context: 128k tokens

Open in the VRAM calculator

Cheapest servers that run it

BF16 weights · 8k context · 8 concurrent requests · about 27 GB

NVIDIA GeForce RTX 5090 In stock

32 GB GDDR7 · Single GPU

$449/mo
NVIDIA L40S In stock

48 GB GDDR6 · Single GPU

$890/mo
NVIDIA RTX PRO 6000 Blackwell In stock

96 GB GDDR7 · Single GPU

$1,390/mo

Delivered in under 10 minutes after payment.

Inference memory

How much VRAM Llama 3.1 8B needs

Weights take parameters × bytes per parameter. Serving adds the KV cache and about 10% runtime overhead; the totals below assume an 8k context with 8 concurrent requests and a BF16 KV cache.

Llama 3.1 8B memory by precision and the cheapest server that fits
Precision	Weights	Total to serve	Cheapest server that fits
BF16 2 bytes per parameter	16 GB	27 GB	NVIDIA GeForce RTX 5090 $449/mo
FP8 1 byte per parameter	8.0 GB	18 GB	NVIDIA GeForce RTX 5090 $449/mo
4-bit 0.5 bytes per parameter	4.0 GB	14 GB	NVIDIA GeForce RTX 5090 $449/mo

GGUF quantizations for llama.cpp and Ollama

Llama 3.1 8B GGUF file sizes and the smallest GPU for one 8k request
Quantization	Weights	Smallest GPU, one 8k request
Q8_0 about 8.5 bits per weight	8.5 GB	NVIDIA GeForce RTX 5090 32 GB · $449/mo
Q6_K about 6.56 bits per weight	6.6 GB	NVIDIA GeForce RTX 5090 32 GB · $449/mo
Q5_K_M about 5.69 bits per weight	5.7 GB	NVIDIA GeForce RTX 5090 32 GB · $449/mo
Q4_K_M about 4.85 bits per weight	4.9 GB	NVIDIA GeForce RTX 5090 32 GB · $449/mo

Estimates with about 8% of GPU memory kept free. Quantized weights also keep some layers in higher precision, and frameworks differ in overhead; check the free memory your engine reports at startup. A model split across GPUs with tensor parallelism runs best on NVLink servers (HGX); the 8× RTX PRO 6000 server links its GPUs over PCIe.

GPUs and servers

Which GPUs can run Llama 3.1 8B?

Yes. In BF16, a single NVIDIA GeForce RTX 5090 (32 GB, $449 a month at OverGPU) runs Llama 3.1 8B with an 8k context and 8 concurrent requests. The table checks every configuration we rent in each precision.

Whether each GPU configuration runs Llama 3.1 8B in BF16, FP8 and 4-bit
Configuration	BF16	FP8	4-bit
NVIDIA GeForce RTX 5090 32 GB · $449/mo	Yes	Yes	Yes
NVIDIA L40S 48 GB · $890/mo	Yes	Yes	Yes
NVIDIA H100 SXM 80 GB · $2,190/mo	Yes	Yes	Yes
NVIDIA RTX PRO 6000 Blackwell 96 GB · $1,390/mo	Yes	Yes	Yes
NVIDIA H200 141 GB · $3,090/mo	Yes	Yes	Yes
NVIDIA B200 180 GB · $4,590/mo	Yes	Yes	Yes
NVIDIA B300 288 GB · $5,190/mo	Yes	Yes	Yes
8× NVIDIA HGX H100 8 × 80 GB · $16,900/mo	Yes	Yes	Yes
8× NVIDIA RTX PRO 6000 Blackwell 8 × 96 GB · $10,490/mo	Yes	Yes	Yes
8× NVIDIA HGX H200 8 × 141 GB · $23,900/mo	Yes	Yes	Yes
8× NVIDIA HGX B200 8 × 180 GB · $34,900/mo	Yes	Yes	Yes
8× NVIDIA HGX B300 8 × 288 GB · $39,900/mo	Yes	Yes	Yes
8× AMD Instinct MI355X 8 × 288 GB · $19,900/mo	Yes	Yes	Yes

Yes: 8 concurrent requests with an 8k context each. 1 req: a single 8k request at a time. —: the weights, KV cache and runtime do not fit. Same formulas as the tables above, with about 8% of GPU memory kept free.

KV cache

Llama 3.1 8B KV cache by context length

Each token in flight stores keys and values for every layer: 2 × 32 layers × 8 KV heads × 128 dimensions × 2 bytes = 128 KB per token in BF16. An FP8 KV cache halves these figures.

KV cache size by context length and number of concurrent requests
Context per request	1 request	8 requests	32 requests
4k tokens	537 MB	4.3 GB	17 GB
32k tokens	4.3 GB	34 GB	137 GB
128k tokens	17 GB	137 GB	550 GB

Fine-tuning

Memory to fine-tune Llama 3.1 8B

Rough totals before tuning batch size and sequence length: QLoRA keeps 4-bit base weights, LoRA keeps BF16 weights frozen, and a full fine-tune with AdamW needs about 16 bytes per parameter plus activations, sharded across GPUs.

Fine-tuning memory by method and the cheapest server that fits
Method	Approximate memory	Cheapest server that fits
QLoRA (4-bit base)	6.1 GB	NVIDIA GeForce RTX 5090 $449/mo
LoRA (BF16 base)	20 GB	NVIDIA GeForce RTX 5090 $449/mo
Full fine-tune (AdamW, mixed precision)	154 GB	NVIDIA B200 $4,590/mo

Full fine-tuning assumes sharded training (FSDP or ZeRO-3) with activation checkpointing. See the fine-tuning guide for methods and settings.

About Llama 3.1 8B

Meta’s Llama 3.1 8B is the small dense model many teams start with: 32 layers, grouped-query attention with 8 KV heads and a 128k-token context window. It is small enough to serve in BF16 on a single card, so the real sizing question is how much KV cache you need for your context length and concurrency.

Every figure on this page comes from the model’s published architecture and the same formulas as our VRAM calculator, where you can change the context length, concurrency and KV cache precision.

Model facts

Developer: Meta
License: Llama 3.1 Community License
Type: Dense
Parameters: 8B
Layers: 32
Attention: 32 query heads, 8 KV headsgrouped-query attention, head dimension 128
Context window: 128k tokens

FAQ

Llama 3.1 8B GPU FAQ

The short version. The help center has the details, and customers with an active server get answers by ticket, 24/7.

Visit the help center

How much VRAM does Llama 3.1 8B need?

About 16 GB for the weights in BF16, 8.0 GB in FP8 and 4.0 GB in 4-bit, plus the KV cache: about 131 MB per 1,000 tokens in flight with a BF16 cache. Serving it in BF16 with an 8k context and 8 concurrent requests takes about 27 GB in total.

Can Llama 3.1 8B run on a single GPU?

Yes. In BF16, a single NVIDIA GeForce RTX 5090 (32 GB) serves it with an 8k context and 8 concurrent requests. Longer contexts or more users need more memory for the KV cache, so check your case in the VRAM calculator.

What is the cheapest OverGPU server for Llama 3.1 8B?

For BF16 serving with an 8k context and 8 concurrent requests, the NVIDIA GeForce RTX 5090 at $449 per month is the lowest-priced configuration that fits. For other contexts and concurrency, size it in the VRAM calculator.

How much memory do I need to fine-tune Llama 3.1 8B?

Roughly 6.1 GB with QLoRA, 20 GB with LoRA on BF16 weights, and 154 GB for a full fine-tune with AdamW in mixed precision, before tuning batch size and sequence length. Full fine-tunes of large models are sharded across GPUs with FSDP or ZeRO-3.

Deploy a GPU server

Run Llama 3.1 8B on your own GPUs.

Choose a configuration, create your account in the same form and pay the first month from your balance. Your server is delivered in under 10 minutes.

Delivered in under 10 minutes
Cancel any month
Pay with crypto, up to +50% bonus

Configure RTX 5090 See all prices