SmolVLM Instruct — Hardware Requirements & GPU Compatibility
VisionSmolVLM-Instruct is Hugging Face's compact multimodal model with 2.2 billion parameters, taking arbitrary sequences of images and text as input to return text responses. It handles answering questions about pictures, describing visual scenes, writing stories based on multiple images, or functioning as a standard language model when no visuals are provided. Built upon the Idefics3 architecture, it combines a SmolLM2-1.7B language model with a SigLIP vision encoder, shrinking each 384x384 image patch into 81 visual tokens to lower memory needs. Given its size, this model runs on a single consumer GPU even without heavy quantization. Context capacity reaches 16,384 tokens. Released under the Apache 2.0 license for free commercial and research use, it came out in November 2024. As the biggest model in the original SmolVLM series, it was later joined by smaller options like the 500M and 256M versions, which employ more intense image compression techniques.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 16K)Best on 24 GB (Q4_K_M · ~320 tok/s · 16K)Macs from 8 GB (Q4_K_M · ~35 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run SmolVLM Instruct
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/ggml-org/SmolVLM-Instruct-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 19.1 GB free · 16K context
Download: SmolVLM-Instruct-Q4_K_M.gguf1.11 GBggml-org
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 3060 12GB (7 GB headroom) · Q4_K_M
- Generation speed
- ~114 tok/s
- generation speed
- Cost per 1M output tokens
- $0.15
- per 1M output tokens
How Much VRAM Does SmolVLM Instruct Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 1.7 GB | — | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 16K context · ~395 tok/s | |
| 1.7 GB | — | 3-bit small — 3-bit small quantization | Fits: 16K context · ~388 tok/s | |
| 1.8 GB | — | 3-bit medium — 3-bit medium quantization | Fits: 16K context · ~364 tok/s | |
| 1.8 GB | — | 4-bit · legacy — 4-bit legacy quantization | Fits: 16K context · ~358 tok/s | |
| 2.0 GB | — | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 16K context · ~320 tok/s | |
| 2.3 GB | — | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 16K context · ~285 tok/s | |
| 2.6 GB | — | 6-bit, very good — 6-bit quantization, very good quality | Fits: 16K context · ~256 tok/s | |
| 3.0 GB | — | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 16K context · ~222 tok/s |
Selected Q4_K_M
4-bit medium quantization4.9GBat 16K context
- Weights
- 1.4 GB
- KV cache
- 3.2 GB
- Overhead
- 0.3 GB
Which Hardware Can Run SmolVLM Instruct?
Q4_K_M · 2.0 GBSmolVLM Instruct (Q4_K_M) requires 2.0 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 3+ GB is recommended. Using the full 16K context window can add up to 2.8 GB, bringing total usage to 4.9 GB. 114 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 185 devices with unified memory can run SmolVLM Instruct, including Mac Studio M5 Ultra (256 GB), Mac Studio M5 Ultra (512 GB).
12 GB
12 GB tier →fits · 7.1 GB free · up to 16K ctx
16 GPUs · 1 device family
RTX 3080 Ti~289Simulate output on RTX 3080 TiRTX 3080 12GB~289Simulate output on RTX 3080 12GBRTX 5070~213Simulate output on RTX 5070iPhone~26Simulate output on Apple iPhone 17 Pro20 GB
fits · 15.1 GB free · up to 16K ctx
1 GPU
RX 7900 XT~234Simulate output on RX 7900 XT32–40 GB
fits · 27.1 GB free · up to 16K ctx
1 GPU · 3 Mac classes · 9 device families
RTX 5090~568Simulate output on RTX 5090M5 Max*36 GB~157Simulate output on Mac Studio M5 Max (36 GB)M632 GB~58Simulate output on iMac 24" M6 (32 GB)RTX Spark N1X~95Simulate output on NVIDIA RTX Spark N1X Laptop (32 GB)Orin~65Simulate output on NVIDIA Jetson AGX Orin 32GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| LFM3 Omni 2BLiquid AI | — | 2.4 GB | ~385 | same size |
| Qwen3.5 2BAlibaba | — | 1.9 GB | ~370 | same size |
| Qwen3 VL 2B InstructAlibaba | — | 3.5 GB | ~362 | same size |
| Qwen2 VL 2B InstructAlibaba | — | 2.1 GB | ~390 | same size |
| Kimi K3 DSparkRadixArk | — | 2.2 GB | ~381 | same size |
| Gemma 4 31B IT Speculator.eagle3RedHatAI | — | 5 GB | ~205 | same size |
Fine-tunes & variants(1)
Specifications
- Publisher
- Hugging Face
- Parameters
- 2.2B
- Architecture
- Idefics3ForConditionalGeneration
- Context Length
- 16,384 tokens
- Vocabulary Size
- 49,155
- Release Date
- 2024-11-18
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does SmolVLM Instruct need?
SmolVLM Instruct requires 2.0 GB of VRAM at Q4_K_M, or 5.2 GB at BF16. Full 16K context adds up to 2.8 GB (4.9 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 2.2B × 4.8 bits ÷ 8 = 1.3 GB
KV Cache + Overhead ≈ 0.7 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 3.6 GB (at full 16K context)
VRAM usage by quantization
What's the best quantization for SmolVLM Instruct?
For SmolVLM Instruct, Q4_K_M (2.0 GB) offers the best balance of quality and VRAM usage. Q5_K_S (2.3 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 1.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run SmolVLM Instruct on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 158 other Macs can run SmolVLM Instruct. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 2.0 GB of usable unified memory (RAM minus macOS overhead).
Can I run SmolVLM Instruct locally?
Yes — SmolVLM Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is SmolVLM Instruct?
At Q4_K_M, SmolVLM Instruct can reach ~2342 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~320 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 2.05 × 0.65 = ~2537 tok/s
Estimated speed at Q4_K_M (2.0 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of SmolVLM Instruct?
At Q4_K_M, the download is about 1.35 GB. The full-precision BF16 version is 4.49 GB. The smallest option (IQ2_XXS) is 0.62 GB.
Which GPUs can run SmolVLM Instruct?
114 consumer GPUs can run SmolVLM Instruct at Q4_K_M. Top options include AMD Radeon RX 5600 XT, AMD Radeon RX 5700 XT, AMD Radeon RX 580 8GB. 114 GPUs have plenty of headroom for comfortable inference.
Which devices can run SmolVLM Instruct?
185 devices with unified memory can run SmolVLM Instruct at Q4_K_M (2.0 GB), including AMD Ryzen 7 8845HS (Hawk Point) Laptop, AMD Ryzen AI 7 350 (Krackan Point) Laptop, AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.