Meta
ChatCodingAI codingReasoning3.21B
Chat

Llama 3.2 3B Instruct

Llama · Llama 3.2 Community License

Ultra-lightweight model that runs on virtually any GPU. Surprisingly capable for its size: good at summarization, simple coding tasks, and quick chat. The default choice when speed matters more than depth.

Parameters
3.21B
Architecture
Dense
Context
131,072 tokens
Released
2024-09-25
Engines
llama.cpp, ollama, vLLM, TGI
Builder Tools
Cursor, Continue, Ollama, LM Studio, Open WebUI

Parameters

3.21B

VRAM

2.8 GB

Context

128K

Formats

5

GPUs

43

Llama 3.2 3B Instruct (3.21B) requires 2.8 GB VRAM at recommended quality (Q6_K). At efficient quality (Q4_K_M), it fits in 2.1 GB VRAM, making it compatible with the NVIDIA RTX 4060 Laptop (40-60W). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 650 tok/s at Q8_0. For the best experience, Starter AI Desktop ($582) is recommended.

Source: OwnRig methodology

VRAM (Recommended)

2.8 GB

Quantization

Q6_K

File Size

2.4 GB

Max Context

128K tokens

Primary Use

Chat

Memory

VRAM Requirements

QualityQuantizationVRAMFile Size
fullQ8_03.7 GB3.2 GB
recommendedQ6_K2.8 GB2.4 GB
recommendedQ5_K_M2.5 GB2.1 GB
efficientQ4_K_M2.1 GB1.8 GB
compressedQ3_K_M1.7 GB1.4 GB
Scaling

Context Length Impact

KV cache VRAM at Q6_K quality. Longer context = more memory.

ContextKV CacheTotal VRAM
2K102 MB2.9 GB
4K102 MB2.9 GB
8K307 MB3.1 GB
16K512 MB3.3 GB
32K1 GB3.8 GB
64K2 GB4.8 GB
128K4.1 GB6.9 GB

Compatible GPUs

43 devices
NVIDIA Grace Blackwell Ultra GB300Q8_0650 tok/sExcellent
Apple M4 (16GB Unified)Q8_030 tok/sExcellent
Apple M4 Max (128GB Unified)Q8_0100 tok/sExcellent
Apple M4 Max (36GB Unified)Q8_0100 tok/sExcellent
Apple M4 Max (64GB Unified)Q8_0100 tok/sExcellent
Apple M4 Pro (24GB Unified)Q8_060 tok/sExcellent
Apple M4 Pro (48GB)Q8_060 tok/sExcellent
Apple M4 Ultra (192GB)Q8_0150 tok/sExcellent
NVIDIA GeForce RTX 3060 12GBQ8_090 tok/sExcellent
NVIDIA GeForce RTX 3080 10GBQ8_0140 tok/sExcellent
NVIDIA GeForce RTX 3090Q8_0150 tok/sExcellent
NVIDIA GeForce RTX 4060 8GBQ8_065 tok/sExcellent
NVIDIA GeForce RTX 4060 Ti 16GBQ8_075 tok/sExcellent
NVIDIA GeForce RTX 4070 SuperQ8_0110 tok/sExcellent
NVIDIA GeForce RTX 4070 Ti 12GBQ8_095 tok/sExcellent
NVIDIA GeForce RTX 4070 Ti SuperQ8_0130 tok/sExcellent
NVIDIA RTX 4080 Laptop (120-150W)Q8_067 tok/sExcellent
NVIDIA GeForce RTX 4080 SuperQ8_0140 tok/sExcellent
NVIDIA GeForce RTX 4090Q8_0170 tok/sExcellent
NVIDIA RTX 4090 Laptop (150-175W)Q8_064 tok/sExcellent
NVIDIA GeForce RTX 5080Q8_0160 tok/sExcellent
NVIDIA GeForce RTX 5090Q8_0200 tok/sExcellent
AMD Radeon Pro W7900Q8_065 tok/sExcellent
NVIDIA RTX PRO 6000 BlackwellQ8_0260 tok/sExcellent
NVIDIA RTX PRO 6000 Blackwell Max-QQ8_0239 tok/sExcellent
Apple M1 Pro (16GB Unified)Q8_034 tok/sExcellent
Apple M2 Pro (16GB Unified)Q8_038 tok/sExcellent
NVIDIA GeForce RTX 5060 8GBQ8_075 tok/sExcellent
NVIDIA GeForce RTX 5060 Ti 16GBQ8_084 tok/sExcellent
Apple M3 Pro (18GB Unified)Q8_035 tok/sGood
NVIDIA RTX 4060 Laptop (40-60W)Q8_039 tok/sGood
NVIDIA RTX 4070 Laptop (80-115W)Q8_046 tok/sGood
AMD Radeon RX 7600Q8_051 tok/sGood
AMD Radeon RX 7900 XTXQ8_0129 tok/sGood
AMD Radeon RX 9070Q8_0132 tok/sGood
Apple M2 (8GB Unified)Q8_014 tok/sGood
Apple M2 (16GB Unified)Q8_014 tok/sGood
Apple M3 (8GB Unified)Q8_016 tok/sGood
Apple M3 (16GB Unified)Q8_016 tok/sGood
AMD Radeon RX 9060 XT 16GBQ8_066 tok/sGood
AMD Radeon RX 9060 XT 8GBQ8_066 tok/sGood
Apple M1 (8GB Unified)Q8_08 tok/sAcceptable
Apple M1 (16GB Unified)Q8_08 tok/sAcceptable

Showing 43 of 43 entries

Builder Context

Llama 3.2 3B Instruct is commonly used with Cursor, Continue, Ollama, LM Studio, Open WebUI. For an AI coding workflow, pair it with an embedding model like nomic-embed-text for local RAG.

Hardware

Recommended Builds

Complete PC builds that can run Llama 3.2 3B Instruct.

FAQ

Frequently Asked Questions

How much VRAM does Llama 3.2 3B Instruct need?
Llama 3.2 3B Instruct requires 2.8 GB VRAM at recommended quality (Q6_K). At lower quality settings, it can fit in as little as 1.7 GB.
What is the best GPU for Llama 3.2 3B Instruct?
The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Llama 3.2 3B Instruct, achieving 650 tok/s at Q8_0 with an excellent rating.
Can I run Llama 3.2 3B Instruct on an RTX 4060 Ti?
Yes. On the NVIDIA GeForce RTX 4060 Ti 16GB, Llama 3.2 3B Instruct runs at 75 tok/s (Q8_0, excellent).
What quantization should I use for Llama 3.2 3B Instruct?
For the best quality, use Q6_K (2.8 GB VRAM). If your GPU has limited VRAM, Q3_K_M (1.7 GB) is the most efficient option with acceptable quality.
Is Llama 3.2 3B Instruct good for coding?
Yes. Llama 3.2 3B Instruct is used with Cursor, Continue, Ollama, LM Studio, Open WebUI for local AI coding. For the best coding experience, pair it with an embedding model for local RAG.

Related Guides

All models

Data confidence: estimated. Source

VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Llama is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.