Can I run Llama 3.1 405B Instruct locally?
Llama 3.1 405B Instruct by meta-llama has 406B parameters. At 4-bit with an 8K context it needs about 236 GB of GPU memory. It fits on 2 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 144 GB | 146 GB | 158 GB | 205 GB |
| 3-bit (Q3_K_M) | usable | 187 GB | 189 GB | 201 GB | 248 GB |
| 4-bit (Q4_K_M) | good, most popular | 234 GB | 236 GB | 247 GB | 295 GB |
| 5-bit (Q5_K_M) | very good | 272 GB | 274 GB | 286 GB | 333 GB |
| 6-bit (Q6_K) | near lossless | 313 GB | 315 GB | 326 GB | 374 GB |
| 8-bit (Q8_0) | effectively lossless | 404 GB | 406 GB | 418 GB | 465 GB |
| 16-bit (F16/BF16) | original weights | 759 GB | 761 GB | 772 GB | 820 GB |
Which GPUs and Macs run Llama 3.1 405B Instruct?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Questions
How much VRAM does Llama 3.1 405B Instruct need?
About 236 GB at 4-bit (Q4_K_M) with an 8K context: 231 GB of weights (estimated), 3.9 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 406 GB.
What is the smallest GPU that runs Llama 3.1 405B Instruct?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the Mac M3 Ultra 512GB (512 GB).
Can I run Llama 3.1 405B Instruct on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M3 Ultra 512GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Llama 4 Maverick 17B 128E Instruct
402B · 228 GB at 4-bitLlama 4 Scout 17B 16E Instruct
109B · 65.2 GB at 4-bitLlama 3.3 70B Instruct
71B · 42.7 GB at 4-bitMiniMax M3
427B · 245 GB at 4-bitGLM 4.5
358B · 207 GB at 4-bitGLM 4.7
358B · 207 GB at 4-bitEmbed this
Free badge and "Will it run on my GPU?" widget for model cards, READMEs and blogs. No cookies, no tracking. All options.
Copy a snippet
Markdown badge
[](https://runsonmy.com/model/llama-3-1-405b-instruct/) Widget (script)
<div class="runsonmy-widget" data-model="llama-3-1-405b-instruct"><a href="https://runsonmy.com/model/llama-3-1-405b-instruct/">Will Llama 3.1 405B Instruct run on my GPU? Check on RunsOnMy</a></div>
<script async src="https://runsonmy.com/embed.js"></script> Widget (iframe)
<iframe src="https://runsonmy.com/embed/m/llama-3-1-405b-instruct/" width="100%" height="220" style="max-width:440px;border:0" loading="lazy" title="Will Llama 3.1 405B Instruct run on my GPU? (RunsOnMy)"></iframe> Model facts
| Hugging Face repo | meta-llama/Llama-3.1-405B-Instruct |
| Parameters | 405,853,388,800 (406B) |
| Layers / hidden size | 126 / 16384 |
| Attention | GQA; 128 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 504.0 KiB |
| Max context | 131,072 tokens |
| License | llama3.1 (gated: accept terms on Hugging Face) |
| Released on HF | Jul 16, 2024 |
| Downloads (30d) / likes | 7,465 / 608 |
Sources: Hugging Face model API · config.json (nvidia/Llama-3.1-405B-Instruct-NVFP4) · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.