Qwen2-72B-Instruct

72.7B parameters, text in, text out, 32K tokens of context. 1 provider serves it.

Verified 2026-10-25 Rank #212 in LLMs · score 88.4 method v1 Sources 3 Downloads 30d 23k · #445 of 619 tracked

Can I run it?44GB at INT4Estimatefits a 64 GB Mac (unified memory)method

Can I use it?tongyi-qianwenCustom license; read the repo LICENSE before commercial hostinglicence file2026-10-24

Do people use it?23kdownloads, 30 days#445 of 619 we track · below the median+2 likes, downloads −1% this week723 likesHugging Face2026-10-24

Worth reading

Picked by rule from 13 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 29 HN pts

Alibaba Qwen2.5-Omni-7B: Open Source End-to-End Multimodal AI Model

alizila.com 2025-03-27 29 HN pts · 3 comments about the Qwen2 family family

Latest · 2 years ago

Qwen2-Math

Qwen2-Math is a series of specialized math language models built upon the Qwen2 LLMs that significantly outperforms the mathematical capabilities of open-source models and even closed-source models.

The model series primarily supports English, with bilingual English and Chinese math models scheduled for release soon.
Evaluation of multi-choice benchmarks including MMLU STEM and GaoKao problems uses a 5-shot setting instead of zero-shot.

qwenlm.github.io 2024-08-08 128 HN pts · 38 comments 5.8k words about the Qwen2 family family

13 best of 13 tracked

The 2 picks above, plus these 11. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleQwen2 LLM Releasedqwenlm.github.io · 2024-06-06 · 261 HN pts · official
CodeLlama.cpp Now Supports Qwen2-VL (Vision Language Model)github.com · 2024-12-14
PaperQwen2 Technical Reportarxiv.org · undated
ArticleAn Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instructhuggingface.co · 2024-06-09 · 16 HN pts
ArticleQwen2.5-Max: The Next Leap in AI Innovationmedium.com · 2025-01-29 · 17 HN pts
ArticleQwen2-72B-Instruct: 72B Instruct-Tuned LLMemergentmind.com · undated
ArticleQwen2 72B Instruct — Hardware Requirements & Compatibility | llmrunllmrun.dev · undated
ArticleQwen2-7B-Instruct with TensorRT-LLM: consistently high tokens/SECinferless.com · 2024-09-05
ArticleQwen2-72B-Instruct: What the Model Actually Does and How It Compares - People & Mediapeopleandmedia.com · 2024-12-26
ArticleQwen2 Technical Report [Quick Review]liner.com · 2024-07-15
ArticleQwen2 Technical Reportexa.ai · 2024-07-15

Show all 11

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

#	Cheapest live row	In /1M	Out /1M	Context	Quant	Seen
1	Featherless	—	—	—	—	Oct 25

All 1 provider →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Served by 1 API provider; 24 quantized builds for local use.

As a service1Featherless
On own hardware24quantized builds
Built on138 fine-tunes · 0 adapters · 5 merges
In apps95Hugging Face Spaces

Top Spaces running it: allenai/WildBench ♥ 232 · allenai/ZebraLogic ♥ 95 · vilarin/Translation-Agent-WebUI ♥ 28 · CZLC/BenCzechMark ♥ 26

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 64 GB Mac (unified memory) at INT4

About 44 GB at INT4, 87 GB at FP8, 175 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

7 dated articles and 25 repo commits so far

Teal is the weights (2024-05-28), amber triangles are 7 dated articles (each links to it), grey ticks are repo commits.

What people built on it

37 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

37 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF qwp4w3hyb/Qwen2-72B-Instruct-iMat-GGUF for llama.cpp, Ollama, LM Studio 145.7k downloads MLX heni86/magnum-v2-72b_mlx-4bpw for Apple silicon 181 downloads AWQ Qwen/Qwen2-72B-Instruct-AWQ for vLLM, SGLang 432k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

16 rows

Benchmarks 16 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (16)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
author-reported AlignBench	8.27	Chinese	model author	2026-10-07
author-reported Arena-Hard · Chat and follow instructions	48.1 percent	English	model author	2026-10-07
author-reported C-Eval · Work across languages	83.8 percent	Chinese	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	42.4 percent	English	model author	2026-10-07
author-reported GSM8K · Reason through hard problems	91.1 percent	Mathematics	model author	2026-10-07
author-reported HumanEval · Write code with agents	86 percent	Coding	model author	2026-10-07
author-reported HumanEval · Write code with agents	79 percent	Coding	model author	2026-10-07
author-reported IFEval · Chat and follow instructions	77.6 percent	English	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	35.7 percent	Coding	model author	2026-10-07
author-reported MATH-500 · Reason through hard problems	59.7 percent	Mathematics	model author	2026-10-07
author-reported MBPP · Write code with agents	80.2 percent	Coding	model author	2026-10-07
author-reported MMLU · Reason through hard problems	82.3 percent	English	model author	2026-10-07
author-reported MMLU-Pro · Reason through hard problems	64.4 percent	English	model author	2026-10-07
author-reported MT-Bench · Chat and follow instructions	9.12	English	model author	2026-10-07
author-reported MultiPL-E · Write code with agents	69.2 percent	Coding	model author	2026-10-07
author-reported TheroemQA	44.4 percent	English	model author	2026-10-07

What the card says

2 cited sentences

Summary generated, cited

Qwen2-72B-Instruct is an instruction-tuned decoder language model with 72.7B parameters.^[1]^[2] The model uses the Qwen2ForCausalLM architecture and supports a context length of up to 131,072 tokens.^[3]^[4]

parameters human
“instruction-tuned 72B Qwen2 model”
architectures
“supports a context length of up to 131,072 tokens”

Generated 2026-10-07 15:04 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

5 use cases

What it is good at 5 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Chat and follow instructionsauthor-reportedAuthor reports MT-Bench 9.12 Author reports Arena-Hard 48.1 Author reports IFEval (Prompt Strict-Acc.) 77.6 +4 more
Reason through hard problemsauthor-reportedAuthor reports MMLU 82.3 Author reports MMLU-Pro 64.4 Author reports GPQA 42.4 +2 more
Write code with agentsauthor-reportedAuthor reports HumanEval 86.0 Author reports MBPP 80.2 Author reports MultiPL-E 69.2 +1 more
Work across languagesauthor-reportedAuthor reports C-Eval 83.8
Read long documentsstructuraldecision-model estimate p=1.00

Facts

22 fields

Listed in LLMs

Facts

Qwen2-72B-Instruct has 72.7B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 32,768 tokens (32K). (config.json max_position_embeddings, observed 2026-10-24)
License: tongyi-qianwen (friction: review). Custom license; read the repo LICENSE before commercial hosting. (license file, observed 2026-10-24)
Created 2024-05-28, last modified 2024-10-08. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	Qwen2ForCausalLM
Model type	qwen2
Dtype	bfloat16
Params by dtype	BF16: 72,706,203,648
Size band	XL
MoE	no
Context length	32,768 (config.json max_position_embeddings)
Languages	en
Papers	arXiv:2309.00071
License	tongyi-qianwen (other)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Qwen2-72B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	175 GB · Multi-GPU node (<640 GB)
FP8 (1 byte/param)	87 GB · Multi-GPU node (<640 GB)
INT4 (0.5 bytes/param)	44 GB · Single high-end GPU (<80 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

37 variants

Lineage

Fine-tuned from Qwen/Qwen2-72B (same organization) Family: Qwen/Qwen2 37 derivatives on the Hub

Base model: Qwen/Qwen2-72B

quantized: 24finetune: 8merge: 5

Top variants (37 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
Qwen/Qwen2-72B-Instruct-AWQ	quantized	AWQ	73.0B	XL	other	Qwen	45	432k
Qwen/Qwen2-72B-Instruct-GPTQ-Int4	quantized	GPTQ	73.0B	XL	other	Qwen	37	251.2k
qwp4w3hyb/Qwen2-72B-Instruct-iMat-GGUF	quantized	GGUF	—	—	other	qwp4w3hyb	0	145.7k
Qwen/Qwen2-72B-Instruct-GPTQ-Int8	quantized	GPTQ	73.0B	XL	other	Qwen	20	107.8k
MaziyarPanahi/calme-2.1-qwen2-72b	finetune	—	72.7B	XL	other	MaziyarPanahi	28	51.8k
anthracite-org/magnum-v1-72b	finetune	—	72.7B	XL	other	anthracite-org	170	39.9k
legraphista/Qwen2-72B-Instruct-IMat-GGUF	quantized	GGUF	—	—	other	legraphista	1	43.3k
anthracite-org/magnum-v2-72b	finetune	—	72.7B	XL	other	anthracite-org	40	17.8k
dfurman/Qwen2-72B-Orpo-v0.1	finetune	—	72.7B	XL	other	dfurman	4	19.2k
tomasonjo/text2cypher-qwen2-72b-4bit-gguf	quantized	-4bit	—	—	apache-2.0	tomasonjo	7	18.4k
Undi95/MG-FinalMix-72B	merge	—	72.7B	XL	other	Undi95	6	15.8k
Undi95/MG-FinalMix-72B-GGUF	merge	GGUF	—	—	—	Undi95	2	12.7k
mradermacher/Qwen2-72B-Instruct-i1-GGUF	quantized	GGUF	—	—	other	mradermacher	1	10.4k
second-state/Qwen2-72B-Instruct-GGUF	quantized	GGUF	—	—	apache-2.0	second-state	0	6.8k
MarsupialAI/MG-FinalMix-72B_iMatrix_GGUF	merge	iMatrix	—	—	other	MarsupialAI	1	5.8k
mradermacher/Qwen2-72B-Instruct-GGUF	quantized	GGUF	—	—	other	mradermacher	1	5k
gaianet/Qwen2-72B-Instruct-GGUF	quantized	GGUF	—	—	apache-2.0	gaianet	0	3.6k
featherless-ai-quants/Qwen-Qwen2-72B-Instruct-GGUF	quantized	GGUF	—	—	—	featherless-ai-quants	0	1.7k
DevQuasar-4/Qwen2-72B-Instruct-GGUF	quantized	GGUF	—	—	other	DevQuasar-4	1	1.4k
kowlick/Qwen2-72B-Instruct-Q4_K_M-GGUF	quantized	GGUF	—	—	other	kowlick	1	1.3k
Sri-Vigneshwar-DJ/hawky-ai-Qwen2-Math-72B-Instruct-GGUF	quantized	GGUF	—	—	other	Sri-Vigneshwar-DJ	1	468
Delta-Vector/Mag-Picaro-72B	finetune	—	72.7B	XL	—	Delta-Vector	5	95
Sri-Vigneshwar-DJ/Phi-3-mini-4k-instruct-marketing-content-gguf	quantized	gguf	—	—	other	Sri-Vigneshwar-DJ	1	274
heni86/magnum-v2-72b_mlx-4bpw	quantized	4bpw	11.4B	L	other	heni86	1	181
Model-SafeTensors/magnum-v2-72b	finetune	—	72.7B	XL	other	Model-SafeTensors	0	207
Infermatic/magnum-v2-72b-FP8-Dynamic	quantized	-FP8	72.7B	XL	other	Infermatic	2	82
Delta-Vector/Picaro-72B	merge	—	72.7B	XL	—	Delta-Vector	0	107
kowlick/Qwen2-72B-Instruct-Q4_K_S-GGUF	quantized	GGUF	—	—	other	kowlick	0	86
FuturisticVibes/magnum-v2-72b-8.0bpw-h8-exl2	quantized	0bpw	—	—	other	FuturisticVibes	0	61
FuturisticVibes/magnum-v1-72b-8.0bpw-h8-exl2	quantized	0bpw	—	—	other	FuturisticVibes	0	58
zeromodels/qwen2-72b-instruct	finetune	—	—	—	other	zeromodels	0	53
Enfint/Qwen2-72B-Instruct-GPTQ-Int4	quantized	GPTQ	73.0B	XL	other	Enfint	0	44
mlc-ai/Qwen2-72B-Instruct-q4f16_1-MLC	quantized	—	—	—	—	mlc-ai	0	32
mlc-ai/Qwen2-72B-Instruct-q0f16-MLC	quantized	—	—	—	—	mlc-ai	0	27
mlc-ai/Qwen2-72B-Instruct-q4f32_1-MLC	quantized	—	—	—	—	mlc-ai	0	18
123some/Llama-3.3-70B-Instruct-underwriting	finetune	—	—	—	—	123some	0	0
Abigail45/Nina-Dolphin	merge	—	—	—	apache-2.0	Abigail45	0	0

Repository changelog

25 commits

Repository changelog 25 commits

Date	Commit	SHA
2024-10-08	Update README.md	c867f76
2024-08-21	Add base_model metadata (#22)	fddbbd7
2024-06-06	update README.md	1af63c6
2024-06-06	Merge branch 'main' of hf.co:Qwen/Qwen2-72B-Instruct	abc8073
2024-06-06	update README.md	7c7b419
2024-06-06	Update README.md	afaaf53
2024-06-06	Update README.md	dda2c6b
2024-06-06	Update README.md	30765d9
2024-06-06	Update config.json	32bc2d2
2024-06-03	Update README.md	890c8dc
2024-05-31	Update README.md	04c481c
2024-05-31	Create LICENSE	9213691
2024-05-31	Update README.md	7b5d46f
2024-05-30	update README.md	54e5483
2024-05-30	Merge branch 'main' of hf.co:Qwen/Qwen2-72B-Instruct	f346172
2024-05-30	update README	12354c2
2024-05-30	Update README.md	8874497
2024-05-30	update tokenizer_config.json	36c9f7f
2024-05-30	Update README.md	f5b479e
2024-05-30	Update README.md	ca5c209
2024-05-30	Update README.md	f8f33b8
2024-05-30	Create README.md	1b5c61b
2024-05-30	Create README.md	3ca26dc
2024-05-28	initial commit	1a52a25
2024-05-28	initial commit	c16e09f

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for Qwen2-72B-Instruct, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Featherless hf	Qwen/Qwen2-72B-Instruct	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

13 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official3

Qwen2 LLM Released

qwenlm.github.io2024-06-06officialauthorityHN front pageabout the Qwen2 familyHN 261 pts · 130 comments4,713 words

Alibaba releases Qwen2, a new series of open-source models featuring five sizes ranging from 0.5B to 72B parameters. The update applies Group Query Attention to all sizes and trains on data in 27 additional languages besides English and Chinese. Specific instruction-tuned versions like Qwen2-7B-Instruct and Qwen2-72B-Instruct support extended context lengths up to 128K tokens, while smaller models adopt the Apache 2.0 license and the largest retains the Qianwen License.written, reviewed, cited

Qwen2-57B-A14B disables tie embedding because large sparse embeddings consume a large proportion of total model parameters.
Instruction-tuned models require the Needle in a Haystack task to verify correct long-context understanding beyond perplexity evaluation.
The Qwen2-72B-Instruct model performs comparably to GPT-4 in safety tests across four categories of multilingual unsafe queries.

Qwen2-Math

qwenlm.github.io2024-08-08officialauthorityHN front pageabout the Qwen2 familyHN 128 pts · 38 comments5,833 words

Qwen2-Math is a series of specialized math language models built upon the Qwen2 LLMs that significantly outperforms the mathematical capabilities of open-source models and even closed-source models. The largest model, Qwen2-Math-72B-Instruct, is evaluated on benchmarks including GSM8K, Math, and MMLU-STEM using few-shot chain-of-thought prompting. Instruction tuning combines a dense reward signal with a binary signal via Rejection Sampling and reinforcement learning with Group Relative Policy Optimization (GRPO) after SFT.written, reviewed, cited

The model series primarily supports English, with bilingual English and Chinese math models scheduled for release soon.
Evaluation of multi-choice benchmarks including MMLU STEM and GaoKao problems uses a 5-shot setting instead of zero-shot.
Performance reporting for complex competition evaluations includes Greedy, Maj@64, RM@64, and RM@256 metrics.

An Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instruct

huggingface.co2024-06-09authorityabout the Qwen2 familyHN 16 pts · 6 comments3,368 words

Qwen 2 Instruct models operate under an Apache 2.0 license but exhibit unique government-mandated alignment with significantly (>80%) fewer refusals in Chinese than English. The deccp dataset documents these behaviors, showing that responses to sensitive topics often include defensive nationalist propaganda or warnings rather than standard refusals. While the models are competitive on benchmarks, their output varies wildly in tone depending on the language used.written, reviewed, cited

Qwen2-72B-Instruct ranks just below Llama-3-70b-Instruct on the ChatArena Leaderboard several days after release.
The deccp dataset contains approximately 95 refusal questions used to evaluate Qwen2 Instructs censorship behavior.

Analysis7

Qwen2 Technical Report

arxiv.orgauthorityabout the Qwen2 family12,467 words

The Qwen2 series releases foundational and instruction-tuned models spanning 0.5 to 72 billion parameters, including a 57 billion parameter Mixture-of-Experts model where 14 billion are activated per token. All models were pre-trained on over 7 trillion tokens using a vocabulary of 151,643 regular tokens and 3 control tokens. The flagship 72B base model scores 84.2 on MMLU and 89.5 on GSM8K, while the Instruct variant achieves 9.1 on MT-Bench.written, reviewed, cited

The 57B MoE model activates 14 billion parameters per token while utilizing 64 routed experts.
The 57B MoE model uses 8 shared experts and 8 activated experts out of 64 routed experts.
The 57B MoE model was trained on 4.5 trillion tokens, fewer than the 7 trillion used for other sizes.

Alibaba Qwen2.5-Omni-7B: Open Source End-to-End Multimodal AI Model

alizila.com2025-03-27about the Qwen2 familyHN 29 pts · 3 comments

Qwen2.5-Max: The Next Leap in AI Innovation

medium.com2025-01-29about the Qwen2 familyHN 17 pts · 4 comments

Qwen2-72B-Instruct: 72B Instruct-Tuned LLM

emergentmind.com1,196 words

Qwen2 72B Instruct — Hardware Requirements & Compatibility | llmrun

llmrun.dev1,869 words

Qwen2 Technical Report [Quick Review]

liner.com2024-07-15about the Qwen2 family

Qwen2 Technical Report

exa.ai2024-07-15about the Qwen2 family220 words

All coverage tracked for this model (13) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Qwen2 Technical Report	arxiv.org	—	2026-10-07	exa
Qwen2-72B-Instruct: 72B Instruct-Tuned LLM	emergentmind.com	—	2026-10-07	exa
Qwen2 72B Instruct — Hardware Requirements & Compatibility \| llmrun	llmrun.dev	—	2026-10-07	exa
Alibaba Qwen2.5-Omni-7B: Open Source End-to-End Multimodal AI Model	alizila.com	2025-03-27	2026-10-07	hn
Qwen2.5-Max: The Next Leap in AI Innovation	medium.com	2025-01-29	2026-10-07	hn
Qwen2-72B-Instruct: What the Model Actually Does and How It Compares - People & Media	peopleandmedia.com	2024-12-26	2026-10-07	exa
Llama.cpp Now Supports Qwen2-VL (Vision Language Model)	github.com	2024-12-14	2026-10-07	hn
Qwen2-7B-Instruct with TensorRT-LLM: consistently high tokens/SEC	inferless.com	2024-09-05	2026-10-07	hn
Qwen2-Math	qwenlm.github.io official	2024-08-08	2026-10-07	hn
Qwen2 Technical Report [Quick Review]	liner.com	2024-07-15	2026-10-07	exa
Qwen2 Technical Report	exa.ai	2024-07-15	2026-10-07	exa
An Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instruct	huggingface.co	2024-06-09	2026-10-07	hn
Qwen2 LLM Released	qwenlm.github.io official	2024-06-06	2026-10-07	hn

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

YaRN: Efficient Context Window Extension of Large Language Models cited in card

Bowen Peng, Jeffrey Quesnelle, Honglu Fan +1 · 2023-08-31 · PDF · HF paper page (88 upvotes)

Similar models

3 models

Similar models

Llama-3.3-70B-Instruct · 70.6B
same size band (XL), same task (text-generation), more providers
Qwen2.5-72B-Instruct · 72.7B
same size band (XL), same task (text-generation), more providers
Llama-3.1-70B-Instruct · 70.6B
same size band (XL), same task (text-generation), fewer providers

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:36 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:04 UTC · Last verified 2026-10-25

Follow Qwen2-72B-Instruct

Watch this model

/feeds/models/Qwen--Qwen2-72B-Instruct.xml(RSS 2.0) · every change to Qwen2-72B-Instruct
/feeds/families/Qwen--Qwen2.xml(RSS 2.0) · every change across the Qwen2 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.