RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run Qwen2.5 VL 3B Instruct locally?

Qwen2.5 VL 3B Instruct by Qwen has 3.8B parameters. At 4-bit with an 8K context it needs about 2.8 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.

2.8 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
125Kmax 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/Qwen2.5-VL-3B-Instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx
2-bit (Q2_K) •rough2.0 GB2.2 GB3.0 GB
3-bit (Q3_K_M) •usable2.3 GB2.5 GB3.3 GB
4-bit (Q4_K_M) •good, most popular2.6 GB2.8 GB3.6 GB
5-bit (Q5_K_M) •very good2.9 GB3.0 GB3.9 GB
6-bit (Q6_K) •near lossless3.2 GB3.3 GB4.2 GB
8-bit (Q8_0) •effectively lossless3.9 GB4.0 GB4.9 GB
16-bit (F16/BF16) •original weights6.6 GB6.7 GB7.6 GB

Which GPUs and Macs run Qwen2.5 VL 3B 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 GBFitsF166.7 GB58details
RTX 4060 Ti 16GB16 GBFitsF166.7 GB47details
RTX 309024 GBFitsF166.7 GB151details
RTX 409024 GBFitsF166.7 GB163details
RTX 509032 GBFitsF166.7 GB290details
RTX 5070 Ti16 GBFitsF166.7 GB145details
RX 7900 XTX24 GBFitsF166.7 GB155details
Mac M4 Pro 48GB48 GBFitsF166.7 GB44details
Mac M4 Max 128GB128 GBFitsF166.7 GB88details
Mac M3 Ultra 512GB512 GBFitsF166.7 GB133details
DGX Spark128 GBFitsF166.7 GB44details
RTX 407012 GBFitsF166.7 GB82details
RTX 5060 Ti 16GB16 GBFitsF166.7 GB73details
RTX 40608 GBFitsF166.7 GB44details
Arc B58012 GBFitsF166.7 GB74details
RTX 3060 Ti8 GBFitsF166.7 GB73details
RTX 30708 GBFitsF166.7 GB73details
RTX 3080 10GB10 GBFitsF166.7 GB123details
RTX 3080 12GB12 GBFitsF166.7 GB148details
RTX 3090 Ti24 GBFitsF166.7 GB163details
RTX 4060 Ti 8GB8 GBFitsF166.7 GB47details
RTX 4070 Super12 GBFitsF166.7 GB82details
RTX 4070 Ti12 GBFitsF166.7 GB82details
RTX 4070 Ti Super16 GBFitsF166.7 GB109details
RTX 408016 GBFitsF166.7 GB116details
RTX 4080 Super16 GBFitsF166.7 GB119details
RTX 50608 GBFitsF166.7 GB73details
RTX 507012 GBFitsF166.7 GB109details
RTX 508016 GBFitsF166.7 GB155details
2x RTX 30902×24 GBFitsF166.7 GB151details
2x RTX 40902×24 GBFitsF166.7 GB163details
2x RTX 50902×32 GBFitsF166.7 GB290details
RTX A600048 GBFitsF166.7 GB124details
RTX 6000 Ada48 GBFitsF166.7 GB155details
RTX PRO 6000 Blackwell96 GBFitsF166.7 GB290details
L40S48 GBFitsF166.7 GB140details
A100 40GB40 GBFitsF166.7 GB252details
A100 80GB80 GBFitsF166.7 GB330details
H100 PCIe80 GBFitsF166.7 GB324details
H100 SXM80 GBFitsF166.7 GB542details
H200141 GBFitsF166.7 GB777details
RX 76008 GBFitsF166.7 GB47details
RX 7700 XT12 GBFitsF166.7 GB70details
RX 7800 XT16 GBFitsF166.7 GB101details
RX 7900 GRE16 GBFitsF166.7 GB93details
RX 7900 XT20 GBFitsF166.7 GB129details
RX 907016 GBFitsF166.7 GB104details
RX 9070 XT16 GBFitsF166.7 GB104details
Ryzen AI Max+ 395 128GB128 GBFitsF166.7 GB41details
Arc A770 16GB16 GBFitsF166.7 GB91details
Mac M1 16GB16 GBFitsF166.7 GB11details
Mac M2 24GB24 GBFitsF166.7 GB16details
Mac M3 24GB24 GBFitsF166.7 GB16details
Mac M4 16GB16 GBFitsF166.7 GB19details
Mac M4 32GB32 GBFitsF166.7 GB19details
Mac M1 Pro 32GB32 GBFitsF166.7 GB32details
Mac M3 Pro 36GB36 GBFitsF166.7 GB24details
Mac M4 Pro 24GB24 GBFitsF166.7 GB44details
Mac M4 Pro 64GB64 GBFitsF166.7 GB44details
Mac M1 Max 64GB64 GBFitsF166.7 GB65details
Mac M2 Max 96GB96 GBFitsF166.7 GB65details
Mac M3 Max 128GB128 GBFitsF166.7 GB65details
Mac M4 Max 64GB64 GBFitsF166.7 GB88details
Mac M2 Ultra 192GB192 GBFitsF166.7 GB129details
Mac M3 Ultra 256GB256 GBFitsF166.7 GB133details

Questions

How much VRAM does Qwen2.5 VL 3B Instruct need?

About 2.8 GB at 4-bit (Q4_K_M) with an 8K context: 1.8 GB of weights (real GGUF file size), 0.3 GB of KV cache and 0.7 GB of runtime overhead. At 8-bit it needs about 4.0 GB.

What is the smallest GPU that runs Qwen2.5 VL 3B 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 Qwen2.5 VL 3B 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.

Qwen3 4B

4.0B · 4.1 GB at 4-bit

Qwen3 4B Instruct 2507

4.0B · 4.1 GB at 4-bit

Qwen2.5 3B Instruct

3.1B · 2.9 GB at 4-bit

Phi 3 mini 4k instruct

3.8B · 5.7 GB at 4-bit

Phi 3.5 mini instruct

3.8B · 5.7 GB at 4-bit

Phi 3 mini 128k instruct

3.8B · 5.7 GB at 4-bit

Model facts

Hugging Face repoQwen/Qwen2.5-VL-3B-Instruct
Parameters3,754,622,976 (3.8B)
Layers / hidden size36 / 2048
AttentionGQA; 16 heads, 2 KV heads, head dim 128
KV cache per token (FP16)36.0 KiB
Max context128,000 tokens
License—
Released on HFJan 26, 2025
Downloads (30d) / likes1,832,802 / 749

Sources: Hugging Face model API · config.json (Qwen/Qwen2.5-VL-3B-Instruct) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.