RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run SmolVLM Instruct locally?

SmolVLM Instruct by HuggingFaceTB has 2.2B parameters. At 4-bit with an 8K context it needs about 3.0 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.

3.0 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
16Kmax context
MHAattention type

Memory needed by quant and context length

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

QuantQuality4K ctx8K ctx
2-bit (Q2_K)rough2.0 GB2.8 GB
3-bit (Q3_K_M)usable2.3 GB3.0 GB
4-bit (Q4_K_M) •good, most popular2.3 GB3.0 GB
5-bit (Q5_K_M)very good2.7 GB3.5 GB
6-bit (Q6_K)near lossless3.0 GB3.7 GB
8-bit (Q8_0) •effectively lossless3.0 GB3.8 GB
16-bit (F16/BF16) •original weights4.6 GB5.4 GB

Which GPUs and Macs run SmolVLM 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 GBFitsF165.4 GB99details
RTX 4060 Ti 16GB16 GBFitsF165.4 GB79details
RTX 309024 GBFitsF165.4 GB258details
RTX 409024 GBFitsF165.4 GB278details
RTX 509032 GBFitsF165.4 GB494details
RTX 5070 Ti16 GBFitsF165.4 GB247details
RX 7900 XTX24 GBFitsF165.4 GB265details
Mac M4 Pro 48GB48 GBFitsF165.4 GB75details
Mac M4 Max 128GB128 GBFitsF165.4 GB151details
Mac M3 Ultra 512GB512 GBFitsF165.4 GB226details
DGX Spark128 GBFitsF165.4 GB75details
RTX 407012 GBFitsF165.4 GB139details
RTX 5060 Ti 16GB16 GBFitsF165.4 GB124details
RTX 40608 GBFitsF165.4 GB75details
Arc B58012 GBFitsF165.4 GB126details
RTX 3060 Ti8 GBFitsF165.4 GB124details
RTX 30708 GBFitsF165.4 GB124details
RTX 3080 10GB10 GBFitsF165.4 GB210details
RTX 3080 12GB12 GBFitsF165.4 GB252details
RTX 3090 Ti24 GBFitsF165.4 GB278details
RTX 4060 Ti 8GB8 GBFitsF165.4 GB79details
RTX 4070 Super12 GBFitsF165.4 GB139details
RTX 4070 Ti12 GBFitsF165.4 GB139details
RTX 4070 Ti Super16 GBFitsF165.4 GB185details
RTX 408016 GBFitsF165.4 GB198details
RTX 4080 Super16 GBFitsF165.4 GB203details
RTX 50608 GBFitsF165.4 GB124details
RTX 507012 GBFitsF165.4 GB185details
RTX 508016 GBFitsF165.4 GB265details
2x RTX 30902×24 GBFitsF165.4 GB258details
2x RTX 40902×24 GBFitsF165.4 GB278details
2x RTX 50902×32 GBFitsF165.4 GB494details
RTX A600048 GBFitsF165.4 GB212details
RTX 6000 Ada48 GBFitsF165.4 GB265details
RTX PRO 6000 Blackwell96 GBFitsF165.4 GB494details
L40S48 GBFitsF165.4 GB238details
A100 40GB40 GBFitsF165.4 GB429details
A100 80GB80 GBFitsF165.4 GB562details
H100 PCIe80 GBFitsF165.4 GB552details
H100 SXM80 GBFitsF165.4 GB924details
H200141 GBFitsF165.4 GB1324details
RX 76008 GBFitsF165.4 GB79details
RX 7700 XT12 GBFitsF165.4 GB119details
RX 7800 XT16 GBFitsF165.4 GB172details
RX 7900 GRE16 GBFitsF165.4 GB159details
RX 7900 XT20 GBFitsF165.4 GB221details
RX 907016 GBFitsF165.4 GB176details
RX 9070 XT16 GBFitsF165.4 GB176details
Ryzen AI Max+ 395 128GB128 GBFitsF165.4 GB71details
Arc A770 16GB16 GBFitsF165.4 GB154details
Mac M1 16GB16 GBFitsF165.4 GB19details
Mac M2 24GB24 GBFitsF165.4 GB28details
Mac M3 24GB24 GBFitsF165.4 GB28details
Mac M4 16GB16 GBFitsF165.4 GB33details
Mac M4 32GB32 GBFitsF165.4 GB33details
Mac M1 Pro 32GB32 GBFitsF165.4 GB55details
Mac M3 Pro 36GB36 GBFitsF165.4 GB41details
Mac M4 Pro 24GB24 GBFitsF165.4 GB75details
Mac M4 Pro 64GB64 GBFitsF165.4 GB75details
Mac M1 Max 64GB64 GBFitsF165.4 GB110details
Mac M2 Max 96GB96 GBFitsF165.4 GB110details
Mac M3 Max 128GB128 GBFitsF165.4 GB110details
Mac M4 Max 64GB64 GBFitsF165.4 GB151details
Mac M2 Ultra 192GB192 GBFitsF165.4 GB221details
Mac M3 Ultra 256GB256 GBFitsF165.4 GB226details

Questions

How much VRAM does SmolVLM Instruct need?

About 3.0 GB at 4-bit (Q4_K_M) with an 8K context: 1.0 GB of weights (real GGUF file size), 1.5 GB of KV cache and 0.5 GB of runtime overhead. At 8-bit it needs about 3.8 GB.

What is the smallest GPU that runs SmolVLM 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 SmolVLM 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.

Model facts

Hugging Face repoHuggingFaceTB/SmolVLM-Instruct
Parameters2,246,272,880 (2.2B)
Layers / hidden size24 / 2048
AttentionMHA; 32 heads, 32 KV heads, head dim 64
KV cache per token (FP16)192.0 KiB
Max context16,384 tokens
Licenseapache-2.0
Released on HFNov 18, 2024
Downloads (30d) / likes17,117 / 602

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