RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run deepseek vl2 tiny locally?

deepseek vl2 tiny by deepseek-ai has 3.4B parameters (1.2B active per token, 64 experts, 6 used per token). 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
4Kmax context
MHAattention type

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.

QuantQuality4K ctx
2-bit (Q2_K)rough2.1 GB
3-bit (Q3_K_M)usable2.4 GB
4-bit (Q4_K_M)good, most popular2.8 GB
5-bit (Q5_K_M)very good3.1 GB
6-bit (Q6_K)near lossless3.5 GB
8-bit (Q8_0)effectively lossless4.2 GB
16-bit (F16/BF16)original weights7.2 GB

Which GPUs and Macs run deepseek vl2 tiny?

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 GBFitsF167.4 GB153details
RTX 4060 Ti 16GB16 GBFitsF167.4 GB123details
RTX 309024 GBFitsF167.4 GB398details
RTX 409024 GBFitsF167.4 GB429details
RTX 509032 GBFitsF167.4 GB762details
RTX 5070 Ti16 GBFitsF167.4 GB381details
RX 7900 XTX24 GBFitsF167.4 GB408details
Mac M4 Pro 48GB48 GBFitsF167.4 GB116details
Mac M4 Max 128GB128 GBFitsF167.4 GB232details
Mac M3 Ultra 512GB512 GBFitsF167.4 GB348details
DGX Spark128 GBFitsF167.4 GB116details
RTX 407012 GBFitsF167.4 GB214details
RTX 5060 Ti 16GB16 GBFitsF167.4 GB191details
RTX 40608 GBFitsQ8_04.5 GB218details
Arc B58012 GBFitsF167.4 GB194details
RTX 3060 Ti8 GBFitsQ8_04.5 GB359details
RTX 30708 GBFitsQ8_04.5 GB359details
RTX 3080 10GB10 GBFitsF167.4 GB323details
RTX 3080 12GB12 GBFitsF167.4 GB388details
RTX 3090 Ti24 GBFitsF167.4 GB429details
RTX 4060 Ti 8GB8 GBFitsQ8_04.5 GB231details
RTX 4070 Super12 GBFitsF167.4 GB214details
RTX 4070 Ti12 GBFitsF167.4 GB214details
RTX 4070 Ti Super16 GBFitsF167.4 GB286details
RTX 408016 GBFitsF167.4 GB305details
RTX 4080 Super16 GBFitsF167.4 GB313details
RTX 50608 GBFitsQ8_04.5 GB359details
RTX 507012 GBFitsF167.4 GB286details
RTX 508016 GBFitsF167.4 GB408details
2x RTX 30902×24 GBFitsF167.4 GB398details
2x RTX 40902×24 GBFitsF167.4 GB429details
2x RTX 50902×32 GBFitsF167.4 GB762details
RTX A600048 GBFitsF167.4 GB327details
RTX 6000 Ada48 GBFitsF167.4 GB408details
RTX PRO 6000 Blackwell96 GBFitsF167.4 GB762details
L40S48 GBFitsF167.4 GB368details
A100 40GB40 GBFitsF167.4 GB661details
A100 80GB80 GBFitsF167.4 GB867details
H100 PCIe80 GBFitsF167.4 GB851details
H100 SXM80 GBFitsF167.4 GB1425details
H200141 GBFitsF167.4 GB2042details
RX 76008 GBFitsQ8_04.5 GB231details
RX 7700 XT12 GBFitsF167.4 GB184details
RX 7800 XT16 GBFitsF167.4 GB265details
RX 7900 GRE16 GBFitsF167.4 GB245details
RX 7900 XT20 GBFitsF167.4 GB340details
RX 907016 GBFitsF167.4 GB272details
RX 9070 XT16 GBFitsF167.4 GB272details
Ryzen AI Max+ 395 128GB128 GBFitsF167.4 GB109details
Arc A770 16GB16 GBFitsF167.4 GB238details
Mac M1 16GB16 GBFitsF167.4 GB29details
Mac M2 24GB24 GBFitsF167.4 GB43details
Mac M3 24GB24 GBFitsF167.4 GB43details
Mac M4 16GB16 GBFitsF167.4 GB51details
Mac M4 32GB32 GBFitsF167.4 GB51details
Mac M1 Pro 32GB32 GBFitsF167.4 GB85details
Mac M3 Pro 36GB36 GBFitsF167.4 GB64details
Mac M4 Pro 24GB24 GBFitsF167.4 GB116details
Mac M4 Pro 64GB64 GBFitsF167.4 GB116details
Mac M1 Max 64GB64 GBFitsF167.4 GB170details
Mac M2 Max 96GB96 GBFitsF167.4 GB170details
Mac M3 Max 128GB128 GBFitsF167.4 GB170details
Mac M4 Max 64GB64 GBFitsF167.4 GB232details
Mac M2 Ultra 192GB192 GBFitsF167.4 GB340details
Mac M3 Ultra 256GB256 GBFitsF167.4 GB348details

Questions

How much VRAM does deepseek vl2 tiny need?

About 3.0 GB at 4-bit (Q4_K_M) with an 8K context: 1.9 GB of weights (estimated), 0.5 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 4.5 GB.

What is the smallest GPU that runs deepseek vl2 tiny?

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 deepseek vl2 tiny 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 repodeepseek-ai/deepseek-vl2-tiny
Parameters3,370,501,440 (3.4B) · ~1.2B active (computed from config)
Layers / hidden size12 / 1280
AttentionMHA; 10 heads, 10 KV heads, head dim 128
KV cache per token (FP16)60.0 KiB
Max context4,096 tokens
Licenseother
Released on HFDec 13, 2024
Downloads (30d) / likes66,628 / 273

Sources: Hugging Face model API · config.json (deepseek-ai/deepseek-vl2-tiny) · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.