Llama · 32B to 80B
Llama 3.3 70B Instruct
70.55 billion parameters across 80 layers. At four-bit precision the weights alone come to 39.67 GB, before any conversation is loaded. It uses grouped-query attention with 8 KV heads for 64 query heads, which cuts the cache to 13% of what full attention would need.
About Llama 3.3 70B Instruct
Meta's December 2024 refresh of the 70B Llama, text only, under the Llama 3.3 community license. It is the model most people mean when they ask whether a machine can run "a 70B", and at 4 bits it needs 42 GB or more.
Llama 3.3 70B Instruct is a dense model with 70.55B parameters, built from 80 layers with a hidden size of 8192. Its repository was published on 2024-11-26. The license listed on Hugging Face is llama3.3. It uses grouped-query attention with 8 KV heads for 64 query heads, which cuts the cache to 13% of what full attention would need.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 69.81 GB | 2.5 GB | 71.81 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 53.88 GB | 2.5 GB | 56.12 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 46.73 GB | 2.5 GB | 49.08 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 39.67 GB | 2.5 GB | 42.12 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 34.91 GB | 2.5 GB | 37.43 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 32.11 GB | 2.5 GB | 34.68 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 25.13 GB | 2.5 GB | 27.8 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. The cache grows in a straight line with every token in the window.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 1.25 GB | 40.87 GB | no | no | no |
| 8k | 2.5 GB | 42.12 GB | no | no | no |
| 16k | 5 GB | 44.62 GB | no | no | no |
| 32k | 10 GB | 49.62 GB | no | no | no |
| 64k | 20 GB | 59.62 GB | no | no | no |
| 128k | 40 GB | 79.62 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Llama 3.3 70B Instruct
15 of 118 consumer devices run it at a quantization worth using, at an 8k context window. Another 7 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs well | Q8_0 | 71.81 GB | 18.5 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs well | Q4_K_M | 42.12 GB | 17 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs well | Q4_K_M | 42.12 GB | 13.6 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs well | Q5_K_M | 49.74 GB | 11.2 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs well | Q5_K_M | 49.74 GB | 11.2 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q5_K_M | 49.74 GB | 11.2 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs well | Q4_K_M | 42.12 GB | 13 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs well | Q6_K | 56.89 GB | 10 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q6_K | 56.89 GB | 10 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q6_K | 56.89 GB | 10 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Runs well | IQ4_XS | 37.92 GB | 10.1 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs well | IQ4_XS | 37.92 GB | 10.1 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | IQ3_XXS ! | 27.8 GB | 48.4 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Fits, but slow | Q8_0 | 72.82 GB | 3.8 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Fits, but slow | Q8_0 | 72.85 GB | 2.8 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Fits, but slow | Q8_0 | 72.85 GB | 2 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 38.7 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 38.7 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 35.9 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 34.4 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 31.3 t/s |
| RTX A5000 NVIDIA | 24 GB | Runs great | IQ2_XXS ! | 19.71 GB | 29.5 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | n/a ! | 28.14 GB | 13.6 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Just fits | n/a | 49.74 GB | 11.2 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Just fits | n/a | 49.74 GB | 11.2 t/s |
Show 35 more devices
| Apple M4 Max 48GB Apple | 48 GB | Just fits | n/a | 37.92 GB | 10.1 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 10 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 10 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Just fits | n/a ! | 28.14 GB | 10 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Just fits | n/a ! | 19.93 GB | 9.7 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Just fits | n/a | 42.68 GB | 6.5 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Just fits | n/a ! | 28.17 GB | 5.5 t/s |
| Apple M5 24GB Apple | 24 GB | Just fits | n/a ! | 19.93 GB | 5.4 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Just fits | n/a ! | 28.17 GB | 5.3 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 5 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 5 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Just fits | n/a | 56.89 GB | 4.9 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Just fits | n/a | 56.89 GB | 4.9 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Just fits | n/a | 56.89 GB | 4.9 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Just fits | n/a | 42.68 GB | 4.5 t/s |
| Apple M4 24GB Apple | 24 GB | Just fits | n/a ! | 19.93 GB | 4.3 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Just fits | n/a | 72.82 GB | 3.8 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Just fits | n/a | 72.82 GB | 3.8 t/s |
| Apple M5 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 3.8 t/s |
| Ryzen AI 9 HX 370 24GB AMD | 24 GB | Just fits | n/a ! | 19.96 GB | 3.8 t/s |
| Apple M3 Pro 36GB Apple | 36 GB | Just fits | n/a ! | 28.14 GB | 3.7 t/s |
| Apple M2 24GB Apple | 24 GB | Just fits | n/a ! | 19.93 GB | 3.5 t/s |
| Apple M3 24GB Apple | 24 GB | Just fits | n/a ! | 19.93 GB | 3.5 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Just fits | n/a | 56.89 GB | 3.3 t/s |
| Apple M4 32GB Apple | 32 GB | Just fits | n/a ! | 28.14 GB | 3 t/s |
| Intel Core Ultra 9 288V 32GB Intel | 32 GB | Just fits | n/a ! | 28.17 GB | 2.8 t/s |
| Ryzen AI 9 HX 370 32GB AMD | 32 GB | Just fits | n/a ! | 28.17 GB | 2.7 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Just fits | n/a | 56.92 GB | 2.7 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Just fits | n/a | 56.92 GB | 2.6 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Just fits | n/a | 72.85 GB | 2 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Partial offload 35/80 | n/a | 42.12 GB | 1.2 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Partial offload 28/80 | n/a | 42.12 GB | 1.1 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Partial offload 28/80 | n/a | 42.12 GB | 1.1 t/s |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Partial offload 28/80 | n/a | 42.12 GB | 1.1 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Partial offload 28/80 | n/a | 42.12 GB | 1.1 t/s |
Can my GPU run Llama 3.3 70B Instruct?
A direct answer for popular devices, each with the settings we would use.
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| Llama 3.3 70B Instruct | 39.7 GB | does not fit well | does not fit well |
| Llama 3 3 Nemotron Super 49B V1 | 28.0 GB | does not fit well | does not fit well |
| Llama 3.1 8B Instruct | 4.5 GB | 31.4 t/s at Q8_0 | 88 t/s at Q8_0 |
| Llama 3.2 3B Instruct | 1.8 GB | 65.4 t/s at Q8_0 | 183.1 t/s at Q8_0 |
| Qwen3 Coder Next | 44.8 GB | 33.2 t/s at Q4_K_M | 56.4 t/s at Q4_K_M |
| Qwen2.5 72B Instruct | 40.9 GB | does not fit well | does not fit well |
What to buy to run Llama 3.3 70B Instruct
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
RTX A6000
$3,500 used
IQ4_XS · uses 37.43 GB of its 48 GB · about 15.3 tokens/s
Best value
Apple M3 Ultra 96GB
$3,999 whole machine
IQ4_XS · about 15.1 tokens/s · 819 GB/s
If it just has to run
Ryzen AI Max+ 395 64GB
$1,699 whole machine
Q3_K_M · about 4.3 tokens/s. Cheaper because it runs the model at a lower precision.
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| RTX A6000 NVIDIA | $3,500 used | 48 GB | IQ4_XS | 15.3 t/s | Check price |
| Apple M3 Ultra 96GB Apple · whole machine | $3,999 | 96 GB | IQ4_XS | 15.1 t/s | Check price |
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | IQ4_XS | 15.1 t/s | Check price |
| RTX 6000 Ada Generation NVIDIA | $6,800 | 48 GB | Q4_K_M | 17 t/s | Check price |
| RTX PRO 6000 Blackwell NVIDIA | $8,500 | 96 GB | Q8_0 | 18.5 t/s | Check price |
| Apple M3 Ultra 512GB Apple · whole machine | $9,499 | 512 GB | IQ4_XS | 15.1 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
These numbers also cover
Fine-tunes share their base model's architecture, so they need exactly the same memory. If you are looking for one of these, the figures above apply unchanged.
meta-llama/Llama-3.1-70B-Instructnvidia/Llama-3.1-Nemotron-70B-Instruct-HFdeepseek-ai/DeepSeek-R1-Distill-Llama-70Blatam-gpt/Llama-3.1-70B-LatamGPT-SFT-1.0
Where it sits in the catalog
Source: meta-llama/Llama-3.3-70B-Instruct on Hugging Face. Downloaded 549k times in the last month. Published 2024-11-26. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Questions people ask about Llama 3.3 70B Instruct
How much VRAM does Llama 3.3 70B Instruct need?
At Q4_K_M with an 8k context window, Llama 3.3 70B Instruct needs about 42.12 GB: 39.08 GB of weights, 2.5 GB of KV cache and 0.54 GB of overhead. At the highest-quality file it needs about 71.81 GB.
What is the cheapest GPU that runs Llama 3.3 70B Instruct?
The smallest consumer device in our catalog that fits it is the Apple M4 Max 64GB with 64 GB, running it at IQ4_XS at roughly 10.1 tokens per second.
How much does a longer context window add for Llama 3.3 70B Instruct?
On the memory side, a 32k window adds 10.0 GB of KV cache on top of the weights, and the cache grows in a straight line with every token. The longest window the model supports natively is 128k tokens.