RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run Llama 3.1 8B Instruct locally?

Llama 3.1 8B Instruct by meta-llama has 8.0B parameters. At 4-bit with an 8K context it needs about 6.2 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.

6.2 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
128Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Llama-3.1-8B-Instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough4.1 GB4.6 GB7.6 GB19.6 GB
3-bit (Q3_K_M) •usable4.9 GB5.4 GB8.4 GB20.4 GB
4-bit (Q4_K_M) •good, most popular5.7 GB6.2 GB9.2 GB21.2 GB
5-bit (Q5_K_M) •very good6.5 GB7.0 GB10.0 GB22.0 GB
6-bit (Q6_K) •near lossless7.3 GB7.8 GB10.8 GB22.8 GB
8-bit (Q8_0)effectively lossless9.1 GB9.6 GB12.6 GB24.6 GB
16-bit (F16/BF16) •original weights16.1 GB16.6 GB19.6 GB31.6 GB

Which GPUs and Macs run Llama 3.1 8B 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).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsQ8_09.6 GB42details
RTX 4060 Ti 16GB16 GBFitsQ8_09.6 GB34details
RTX 309024 GBFitsF1616.6 GB58details
RTX 409024 GBFitsF1616.6 GB63details
RTX 509032 GBFitsF1616.6 GB112details
RTX 5070 Ti16 GBFitsQ8_09.6 GB105details
RX 7900 XTX24 GBFitsF1616.6 GB60details
Mac M4 Pro 48GB48 GBFitsF1616.6 GB17details
Mac M4 Max 128GB128 GBFitsF1616.6 GB34details
Mac M3 Ultra 512GB512 GBFitsF1616.6 GB51details
DGX Spark128 GBFitsF1616.6 GB17details
RTX 407012 GBFitsQ8_09.6 GB59details
RTX 5060 Ti 16GB16 GBFitsQ8_09.6 GB53details
RTX 40608 GBFitsQ4_K_M6.2 GB55details
Arc B58012 GBFitsQ8_09.6 GB53details
RTX 3060 Ti8 GBFitsQ4_K_M6.2 GB91details
RTX 30708 GBFitsQ4_K_M6.2 GB91details
RTX 3080 10GB10 GBFitsQ6_K7.8 GB115details
RTX 3080 12GB12 GBFitsQ8_09.6 GB107details
RTX 3090 Ti24 GBFitsF1616.6 GB63details
RTX 4060 Ti 8GB8 GBFitsQ4_K_M6.2 GB59details
RTX 4070 Super12 GBFitsQ8_09.6 GB59details
RTX 4070 Ti12 GBFitsQ8_09.6 GB59details
RTX 4070 Ti Super16 GBFitsQ8_09.6 GB79details
RTX 408016 GBFitsQ8_09.6 GB84details
RTX 4080 Super16 GBFitsQ8_09.6 GB86details
RTX 50608 GBFitsQ4_K_M6.2 GB91details
RTX 507012 GBFitsQ8_09.6 GB79details
RTX 508016 GBFitsQ8_09.6 GB113details
2x RTX 30902×24 GBFitsF1616.6 GB58details
2x RTX 40902×24 GBFitsF1616.6 GB63details
2x RTX 50902×32 GBFitsF1616.6 GB112details
RTX A600048 GBFitsF1616.6 GB48details
RTX 6000 Ada48 GBFitsF1616.6 GB60details
RTX PRO 6000 Blackwell96 GBFitsF1616.6 GB112details
L40S48 GBFitsF1616.6 GB54details
A100 40GB40 GBFitsF1616.6 GB97details
A100 80GB80 GBFitsF1616.6 GB127details
H100 PCIe80 GBFitsF1616.6 GB124details
H100 SXM80 GBFitsF1616.6 GB208details
H200141 GBFitsF1616.6 GB299details
RX 76008 GBFitsQ4_K_M6.2 GB59details
RX 7700 XT12 GBFitsQ8_09.6 GB51details
RX 7800 XT16 GBFitsQ8_09.6 GB73details
RX 7900 GRE16 GBFitsQ8_09.6 GB68details
RX 7900 XT20 GBFitsF1616.6 GB50details
RX 907016 GBFitsQ8_09.6 GB75details
RX 9070 XT16 GBFitsQ8_09.6 GB75details
Ryzen AI Max+ 395 128GB128 GBFitsF1616.6 GB16details
Arc A770 16GB16 GBFitsQ8_09.6 GB66details
Mac M1 16GB16 GBFitsQ8_09.6 GB8.0details
Mac M2 24GB24 GBFitsQ8_09.6 GB12details
Mac M3 24GB24 GBFitsQ8_09.6 GB12details
Mac M4 16GB16 GBFitsQ8_09.6 GB14details
Mac M4 32GB32 GBFitsF1616.6 GB7.5details
Mac M1 Pro 32GB32 GBFitsF1616.6 GB12details
Mac M3 Pro 36GB36 GBFitsF1616.6 GB9.3details
Mac M4 Pro 24GB24 GBFitsQ8_09.6 GB32details
Mac M4 Pro 64GB64 GBFitsF1616.6 GB17details
Mac M1 Max 64GB64 GBFitsF1616.6 GB25details
Mac M2 Max 96GB96 GBFitsF1616.6 GB25details
Mac M3 Max 128GB128 GBFitsF1616.6 GB25details
Mac M4 Max 64GB64 GBFitsF1616.6 GB34details
Mac M2 Ultra 192GB192 GBFitsF1616.6 GB50details
Mac M3 Ultra 256GB256 GBFitsF1616.6 GB51details

Prefer an API?

Meta: Llama 3.1 8B Instruct is on OpenRouter at $0.02 per 1M input tokens and $0.04 per 1M output tokens. Price details and history →

Questions

How much VRAM does Llama 3.1 8B Instruct need?

About 6.2 GB at 4-bit (Q4_K_M) with an 8K context: 4.6 GB of weights (real GGUF file size), 1.0 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 9.6 GB.

What is the smallest GPU that runs Llama 3.1 8B Instruct?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).

Can I run Llama 3.1 8B Instruct on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB 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 3.1 8B

8.0B · 6.2 GB at 4-bit

Meta Llama Guard 2 8B

8.0B · 6.2 GB at 4-bit

Llama 3.2 11B Vision

11B · 8.0 GB at 4-bit

Hermes 3 Llama 3.1 8B

8.0B · 6.2 GB at 4-bit

DeepSeek R1 Distill Llama 8B

8.0B · 6.2 GB at 4-bit

Meta Llama 3 8B Instruct

8.0B · 6.2 GB at 4-bit

Model facts

Hugging Face repometa-llama/Llama-3.1-8B-Instruct
Parameters8,030,261,248 (8.0B)
Layers / hidden size32 / 4096
AttentionGQA; 32 heads, 8 KV heads, head dim 128
KV cache per token (FP16)128.0 KiB
Max context131,072 tokens
Licensellama3.1 (gated: accept terms on Hugging Face)
Released on HFJul 18, 2024
Downloads (30d) / likes4,208,415 / 8,965

Sources: Hugging Face model API · config.json (unsloth/Llama-3.1-8B-Instruct-GGUF) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.