# Llama 3.3 70B GPU requirements: VRAM, GPUs and servers

> Llama 3.3 70B needs about 141 GB of VRAM in BF16, 71 GB in FP8 and 35 GB in 4-bit, plus KV cache. GPUs and servers that run it, with prices.

Source: https://overgpu.com/models/llama-3-3-70b  
Updated: 2026-11-03  
Publisher: OverGPU (https://overgpu.com/)

**In short:** Yes. In FP8, a single NVIDIA H200 (141 GB, $3,090 a month at OverGPU) runs Llama 3.3 70B with an 8k context and 8 concurrent requests; one NVIDIA RTX PRO 6000 Blackwell (96 GB) holds a single 8k request.

## Model facts

- **Developer:** Meta
- **License:** Llama 3.3 Community License
- **Type:** Dense
- **Parameters:** 70.6B
- **Layers:** 80
- **Attention:** 64 query heads, 8 KV heads, head dimension 128
- **Context window:** 128k tokens

## How much VRAM Llama 3.3 70B needs

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

| Precision | Weights | Total to serve | Cheapest OverGPU server that fits |
| --- | --- | --- | --- |
| BF16 | 141 GB | 178 GB | NVIDIA B300 ($5,190/month) |
| FP8 | 71 GB | 100 GB | NVIDIA H200 ($3,090/month) |
| 4-bit | 35 GB | 61 GB | NVIDIA RTX PRO 6000 Blackwell ($1,390/month) |

## Which GPUs can run Llama 3.3 70B?

Yes = 8 concurrent requests with an 8k context each; 1 request = a single 8k request at a time; No = does not fit.

| Configuration | BF16 | FP8 | 4-bit |
| --- | --- | --- | --- |
| NVIDIA GeForce RTX 5090 (32 GB, $449/month) | No | No | No |
| NVIDIA L40S (48 GB, $890/month) | No | No | 1 request |
| NVIDIA H100 SXM (80 GB, $2,190/month) | No | No | Yes |
| NVIDIA RTX PRO 6000 Blackwell (96 GB, $1,390/month) | No | 1 request | Yes |
| NVIDIA H200 (141 GB, $3,090/month) | No | Yes | Yes |
| NVIDIA B200 (180 GB, $4,590/month) | 1 request | Yes | Yes |
| NVIDIA B300 (288 GB, $5,190/month) | Yes | Yes | Yes |
| 8× NVIDIA HGX H100 (8 × 80 GB, $16,900/month) | Yes | Yes | Yes |
| 8× NVIDIA RTX PRO 6000 Blackwell (8 × 96 GB, $10,490/month) | Yes | Yes | Yes |
| 8× NVIDIA HGX H200 (8 × 141 GB, $23,900/month) | Yes | Yes | Yes |
| 8× NVIDIA HGX B200 (8 × 180 GB, $34,900/month) | Yes | Yes | Yes |
| 8× NVIDIA HGX B300 (8 × 288 GB, $39,900/month) | Yes | Yes | Yes |
| 8× AMD Instinct MI355X (8 × 288 GB, $19,900/month) | Yes | Yes | Yes |

## Llama 3.3 70B KV cache by context length

About 328 MB per 1,000 tokens in flight with a BF16 cache.

| Context per request | 1 request | 8 requests | 32 requests |
| --- | --- | --- | --- |
| 4k tokens | 1.3 GB | 11 GB | 43 GB |
| 32k tokens | 11 GB | 86 GB | 344 GB |
| 128k tokens | 43 GB | 344 GB | 1,374 GB |

## Memory to fine-tune Llama 3.3 70B

| Method | Approximate memory | Cheapest OverGPU server that fits |
| --- | --- | --- |
| QLoRA (4-bit base) | 53 GB | NVIDIA RTX PRO 6000 Blackwell ($1,390/month) |
| LoRA (BF16 base) | 175 GB | NVIDIA B300 ($5,190/month) |
| Full fine-tune (AdamW, mixed precision) | 1,356 GB | 8× AMD Instinct MI355X ($19,900/month) |

## About Llama 3.3 70B

Llama 3.3 70B is Meta’s 70B dense instruction model: 80 layers, grouped-query attention with 8 KV heads and a 128k-token context window. It is the reference point for 70B-class serving: in BF16 the weights alone take about 141 GB, more than a single H200 can hold once the KV cache is added, while FP8 halves that and fits one H200 with room to spare.

Every figure comes from the model's published architecture and the formulas of the VRAM calculator (https://overgpu.com/tools/vram-calculator).

## Frequently asked questions

### How much VRAM does Llama 3.3 70B need?

About 141 GB for the weights in BF16, 71 GB in FP8 and 35 GB in 4-bit, plus the KV cache: about 328 MB per 1,000 tokens in flight with a BF16 cache. Serving it in FP8 with an 8k context and 8 concurrent requests takes about 100 GB in total.

### Can Llama 3.3 70B run on a single GPU?

Yes. In FP8, a single NVIDIA H200 (141 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 3.3 70B?

For FP8 serving with an 8k context and 8 concurrent requests, the NVIDIA H200 at $3,090 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 3.3 70B?

Roughly 53 GB with QLoRA, 175 GB with LoRA on BF16 weights, and 1,356 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.

## Related pages

- [LLM VRAM calculator: will your model fit on a GPU?](https://overgpu.com/tools/vram-calculator): Pick Llama, Qwen, Mistral or gpt-oss, set the precision and context length, and see weights, KV cache and which GPUs fit.
- [How much VRAM do I need for an LLM?](https://overgpu.com/blog/vram-for-llms): Quick answers for 7B, 14B, 32B, 70B and 405B models at BF16, FP8 and 4-bit, plus KV cache by context and the GPU that fits.
- [Best GPU for LLM inference in 2026: a sizing guide by model size](https://overgpu.com/blog/best-gpu-for-llm-inference): Use our 2026 sizing table, from RTX 5090 to 8× B300, to choose.
- [LLM GPU requirements](https://overgpu.com/models): VRAM, KV cache and the cheapest servers for Llama, Qwen3, Mistral and gpt-oss models.