llmrun
AlibabaQwen 2Qwen2ForCausalLM

Qwen2 72B Instruct — Hardware Requirements & GPU Compatibility

Chat

Qwen2-72B-Instruct is the instruction-tuned, chat-ready 72.7-billion-parameter model from Alibaba's Qwen2 series, built on a Transformer architecture with SwiGLU activation, QKV attention bias, and grouped-query attention, and trained with supervised fine-tuning plus direct preference optimization. It generally surpassed the prior Qwen1.5 line and competed with proprietary models on language understanding, generation, multilingual tasks, coding, and math benchmarks at release, though it has since been superseded by Qwen2.5-72B-Instruct. At 72.7 billion parameters, it needs a multi-GPU workstation to run in full precision. Context length is natively 32,768 tokens, extendable to 131,072 tokens using YaRN scaling, as documented on the model card. It is released under Alibaba's Tongyi Qianwen custom license, which permits commercial use but requires a separate license once a deployment passes 100 million monthly active users. It was published in May 2024.

72.7B params · dense33K contextOther24.0K downloads724 likes3.0K quant downloads

Doesn't fit a 24 GB GPU

Smallest consumer fit → Snapdragon X Elite Copilot+ PC · IQ2_XXS · 25.7 GB · ~4 tok/s

Macs from32 GB IQ2_S · ~12 tok/s

Score 77.5#94 of 106 open#40 of 51 that fit ≤64 GBOther

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

77.5

Benchmarks · tier chart ↓
Among 106 open modelsHow the Composite Score is built

Run Qwen2 72B 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
one command · needs Ollama installed
$ollama run qwen2:72b-instruct-q4_K_M

Q4_K_M · needs ~49.3 GB · 16K context

Download: Qwen2-72B-Instruct-Q4_K_M.gguf47.42 GBsecond-state

install, OS, other quants

Ways to runOther ways

each opens step-by-step

Run in cloud

Your own private copy · Q4_K_M · 96 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

How Much VRAM Does Qwen2 72B Instruct Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn RTX 4090 (reference)
31.9 GB41.9 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsDoes not fit: needs 36.6 GB
32.8 GB42.9 GB3-bit small — 3-bit small quantizationDoes not fit: needs 37.5 GB
36.4 GB46.5 GB3-bit medium — 3-bit medium quantizationDoes not fit: needs 41.1 GB
37.3 GB47.4 GB4-bit · legacy — 4-bit legacy quantizationDoes not fit: needs 42 GB
44.6 GB54.7 GB4-bit medium — 4-bit medium quantization — most popular sweet spotDoes not fit: needs 49.3 GB
52.8 GB62.8 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffDoes not fit: needs 57.5 GB
61.0 GB71 GB6-bit, very good — 6-bit quantization, very good qualityDoes not fit: needs 65.7 GB
73.7 GB83.8 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 78.4 GB
est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

49.3GBat 16K context

Weights
43.6 GB
KV cache
5.4 GB
Overhead
0.3 GB

Doesn't fit RTX 4090 (reference) · needs 25.3 GB more

≥10 tok/s needs ≥683 GB/s · M3 Ultra ~13, M2 Ultra ~13

Or .

Which Hardware Can Run Qwen2 72B Instruct?

Q4_K_M · 44.6 GB

Qwen2 72B Instruct (Q4_K_M) requires 44.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 58+ GB is recommended. Using the full 33K context window can add up to 10.1 GB, bringing total usage to 54.7 GB. No consumer GPU has enough memory. 52 devices with unified memory can run Qwen2 72B Instruct, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Pro M2 Ultra (64 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 25.7 GB8 GB tier →12 GB tier →16 GB tier →
24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 25.7 GB24 GB tier →
Mac 32 GB · Mac Mini M4IQ2_S · 28.4 GB · ~4 tok/s · 16KMacs →
Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 49.3 GB · ~9 tok/s · 32KMacs →

How good is Qwen2 72B Instruct?

The llmrun Score merges 3 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the ≤64 GB tier

  • Qwen2 72B Instruct
  • Other models
  • Best Score for its VRAM

#40 of 51 that fit ≤64 GB

Each dot is a model at the best quant that fits ≤64 GB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤64 GB ranking →

Benchmarks

The 6 boards that weigh most in the llmrun Score, of 7 with a result for this model; badges show who ran the test.

  • GPQA Diamond

    Epoch AI · Jun 2024Independent

    40.8%

    #71/100 open

    Independent
  • LMArena Text

    LMArena · Oct 2026Arena

    1261.9

    #127/188 open

    Arena
  • MMLU-Pro

    TIGER-LabIndependent

    64.4%

    #49/146 open

    Independent
  • MATH Level 5

    Epoch AI · Jun 2024Independent

    39.1%

    #22/39 open

    Independent
  • MMLU

    Epoch AI · Jun 2024Lab-reported

    82.4%

    #8/85 open

    Lab-reported
  • WeirdML

    Epoch AI · Jun 2024Independent

    11.3%

    #44/47 open

    Independent
All 7 Qwen2 72B Instruct benchmarks

In the ≤64 GB class

51 models fit ≤64 GB

Same Score, only models that fit ≤64 GB at their best quant: the ranking a ≤64 GB owner actually chooses from.

  1. #1

    Qwen3.8 27BAlibaba

    122.0

    ~56 tok/s · 17.9 GB

  2. #2

    Qwen3.6 27BAlibaba

    111.2

    ~56 tok/s · 17.9 GB

  3. #3

    Qwen3.5 122B A10BAlibaba

    109.4

    ~62 tok/s · 63.0 GB

  4. #4

    Qwen3.6 35B A3BAlibaba

    106.8

    ~91 tok/s · 22.0 GB

  5. #5

    Gemma 4 31B ITGoogle

    106.7

    ~48 tok/s · 21.7 GB

  6. #6

    Gemma 4 26B A4B ITGoogle

    105.8

    ~114 tok/s · 16.5 GB

Full ≤64 GB ranking →AMD Instinct MI210 page →

Alternatives to Qwen2 72B Instruct

Qwen 2 family · 13 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Similar models

10
ModelScoreVRAMtok/sWhy it is similar
Qwen3.5 122B A10BAlibaba109.475.5 GB—needs > 24 GBsimilar score
GLM 4.5 AirZ.ai98.667.6 GB—needs > 24 GBsimilar score
GPT OSS 120BOpenAI97.570.8 GB—needs > 24 GBsimilar score
NVIDIA Nemotron 3 Super 120B A12B BF16NVIDIA95.874.7 GB—needs > 24 GBsimilar score
Mistral Large Instruct 2411Mistral AI82.579.8 GB—needs > 24 GBsimilar score
Llama 4 Scout 17B 16E InstructMeta82.177 GB—needs > 24 GBsimilar score
Mistral Large Instruct 2407Mistral AI81.786.4 GB—needs > 24 GBsimilar score
C4ai Command A 03 2025Cohere79.878.7 GB—needs > 24 GBsimilar score
Gemma 3 27B ITGoogle79.421.7 GB~36similar score
Mistral Small 3.1 24B Instruct 2503Mistral AI79.018.1 GB~43similar score

Specifications

Publisher
Alibaba
Family
Qwen 2
Parameters
72.7B
Architecture
Qwen2ForCausalLM
Context Length
32,768 tokens
Vocabulary Size
152,064
Release Date
2024-05-28
License
Other

Resources

Frequently Asked Questions

How much VRAM does Qwen2 72B Instruct need?

Qwen2 72B Instruct requires 44.6 GB of VRAM at Q4_K_M, or 146.4 GB at BF16. Full 33K context adds up to 10.1 GB (54.7 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 72.7B × 4.8 bits ÷ 8 = 43.6 GB

KV Cache + Overhead ≈ 1 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 ≈ 11.1 GB (at full 33K context)

VRAM usage by quantization

44.6 GB
54.7 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Qwen2 72B Instruct?

Yes, at IQ2_S (23.7 GB) or lower. Higher quantizations like IQ2_M (25.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Qwen2 72B Instruct?

For Qwen2 72B Instruct, Q4_K_M (44.6 GB) offers the best balance of quality and VRAM usage. Q5_0 (46.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 21.0 GB.

VRAM requirement by quantization

IQ2_XXS
21.0 GB
IQ3_XS
31.0 GB
Q4_0
37.3 GB
Q4_K_M ★
44.6 GB
Q5_0
46.4 GB
BF16
146.4 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Qwen2 72B Instruct on a Mac?

Yes — Mac Studio M1 Max (32 GB) and 84 other Macs can run Qwen2 72B Instruct, the smallest at IQ2_S. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 44.6 GB of usable unified memory (RAM minus macOS overhead), from Mac Pro M2 Ultra (64 GB).

Can I run Qwen2 72B Instruct locally?

Yes — Qwen2 72B Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Qwen2 72B Instruct?

At Q4_K_M, Qwen2 72B Instruct can reach ~108 tok/s on AMD Instinct MI350X. 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 ÷ 44.6 × 0.65 = ~117 tok/s

Estimated speed at Q4_K_M (44.6 GB)

~117 tok/s
~117 tok/s
~108 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Qwen2 72B Instruct?

At Q4_K_M, the download is about 43.62 GB. The full-precision BF16 version is 145.41 GB. The smallest option (IQ2_XXS) is 19.99 GB.

Which GPUs can run Qwen2 72B Instruct?

No single consumer GPU has enough VRAM to run Qwen2 72B Instruct at Q4_K_M (44.6 GB). Multi-GPU or professional hardware is required.

Which devices can run Qwen2 72B Instruct?

52 devices with unified memory can run Qwen2 72B Instruct at Q4_K_M (44.6 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.