Qwen3-4B-Instruct-2507

4.02B parameters, text in, text out, 256K tokens of context. 1 provider serves it from $0.010 per million input tokens.

Verified 2026-10-25 Rank #70 in LLMs · score 104.3 method v1 Sources 3 Downloads 30d 2.6M · #53 of 619 tracked

Changed this week: 13 new quantized builds on the Hub (352 in all).2026-10-24 · 1 new provider: Nscale.2026-10-23 · 1 provider no longer listed: Featherless.2026-10-23

Can I run it?2.4GB at INT4Estimatefits a 16 GB cardmethodCommunity: Qwen3-4B-Instruct-2507 — specs, VRAM, license, benchmarks | LocalLLMs →

What does it cost?$0.010/1M incheapest of 1Nscale2026-10-25

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Does it work?35.68on BFCL#15 of 28 measuredleaderboard2026-10-25

Do people use it?2.6Mdownloads, 30 days#53 of 619 we track · top 9%+8 likes, downloads −12% this week1,022 likesHugging Face2026-10-24

Worth reading

Picked by rule from 18 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 490 HN pts

GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2

OpenAI released gpt-oss-120b and gpt-oss-20b, their first open-weight models since GPT-2 in 2019.

A 2025 research paper confirms Dropout results in worse downstream performance in single-epoch regimes.

magazine.sebastianraschka.com 2025-08-10 490 HN pts · 97 comments 5.2k words about the Qwen3 family family

Latest · 8 months ago

Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers

venturebeat.com 2026-02-28 461 HN pts · 272 comments about the Qwen3 family family

18 best of 18 tracked

The 5 picks above, plus these 13. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleQwen3-Coder-Nextqwen.ai · 2026-02-03 · official
ArticleQwen3-Max-Thinkingqwen.ai · 2026-01-26 · official
ArticleQwen3-TTS family is now open sourced: Voice design, clone, and generationqwen.ai · 2026-01-22 · official
ArticleQwen3: Think deeper, act fasterqwenlm.github.io · 2025-04-28 · 869 HN pts · official
ArticleQwen3-Coder: Agentic coding in the worldqwenlm.github.io · 2025-07-22 · official
ArticleQwen3-Nextqwen.ai · 2025-09-12 · official
ArticleQwen3-VLqwen.ai · 2025-09-23 · official
CodeQwen3-Omni: Native Omni AI model for text, image and videogithub.com · 2025-09-22
Codesrc/qai_hub_models/models/qwen3_4b_instruct_2507/README.mdgithub.com · undated
ArticleQwen3 4B 2507 Instruct | Benched.aibenched.ai · undated
ArticleQwen : Qwen 3 4B Instruct 2507ai-stats.phaseo.app · undated
ArticleQwen3-4B-Instruct-2507: Lightweight Open-Source LLMdev.co · undated
ArticleQwen3 4B 2507 Instruct Review | Pricing, Benchmarks & Capabilities (2026) | Design for Onlinedesignforonline.com · 2026-04-24

Show all 13

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	Nscale	$0.010	$0.030	256K	—	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; heavily run locally (352 quantized builds); a big base for builders (8,506 fine-tunes, adapters and merges).

As a service1Nscale
On own hardware352quantized builds
Built on8,5062,618 fine-tunes · 5,853 adapters · 35 merges
In apps100+Hugging Face Spaces

Top Spaces running it: embedl/hfviewer ♥ 181 · tomg-group-umd/lm-watermarking ♥ 100 · ginigen-ai/smol-worldcup ♥ 59 · dqy08/InfoLens1 ♥ 22

Will it fit?

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

Fits a 16 GB card at INT4

About 2.4 GB at INT4, 4.8 GB at FP8, 9.7 GB at BF16: weights only, so leave room for the KV cache.

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

How it scores

Third-party measurements only; author-reported numbers stay in Reference, marked as such.

#15 of 28 on BFCL so far

35.68. Each tick is a model on this site with a measured BFCL row; author-reported numbers stay in Reference. Percent (overall accuracy). FC

Source · leaderboard · 2026-10-25

The story so far

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

Most coverage came in the first two weeks

Teal is the weights (2025-08-05), amber triangles are 3 dated articles (each links to it), grey ticks are repo commits.

What people built on it

8,858 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

8,858 derivatives on the Hub, most of them adapters

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

GGUF MaziyarPanahi/Qwen3-4B-Instruct-2507-GGUF for llama.cpp, Ollama, LM Studio 1.8M downloads MLX lmstudio-community/Qwen3-4B-Instruct-2507-MLX-4bit for Apple silicon 1.2M downloads FP8 Qwen/Qwen3-4B-Instruct-2507-FP8 for vLLM, SGLang 9M downloads

All variants →

Reference

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

Benchmarks

23 rows

Benchmarks 23 rows

measured third-party leaderboard row (2)author-reported number from the model card or eval files (21)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
measured BFCL · Call tools and APIs	35.68 percent (overall accuracy)	FC 5 sub-scores Non-Live AST Acc: 87.88 Live Acc: 76.39 Multi Turn Acc: 22.12 Web Search Acc: 3 Memory Acc: 17.63	leaderboard	2026-10-25
measured BFCL · Call tools and APIs	35.52 percent (overall accuracy)	Prompt 5 sub-scores Non-Live AST Acc: 86.44 Live Acc: 74.69 Multi Turn Acc: 20.5 Web Search Acc: 4.5 Memory Acc: 23.87	leaderboard	2026-10-25
author-reported AIME 2025 · Reason through hard problems	47.4	—	model author	2026-10-07
author-reported Aider Polyglot · Write code with agents	12.9	—	model author	2026-10-07
author-reported Arena-Hard · Chat and follow instructions	43.4	—	model author	2026-10-07
author-reported BFCL · Call tools and APIs	61.9	—	model author	2026-10-07
author-reported Creative Writing · Chat and follow instructions	83.5	—	model author	2026-10-07
author-reported Creative Writing · Chat and follow instructions	83.4	—	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	62	—	model author	2026-10-07
author-reported HMMT · Reason through hard problems	31	—	model author	2026-10-07
author-reported IFEval · Chat and follow instructions	83.4	—	model author	2026-10-07
author-reported LiveBench · Reason through hard problems	63	—	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	35.1	—	model author	2026-10-07
author-reported MMLU · Reason through hard problems	84.2	—	model author	2026-10-07
author-reported MMLU-Pro · Reason through hard problems	69.6	—	model author	2026-10-07
author-reported MultiPL-E · Write code with agents	76.8	—	model author	2026-10-07
author-reported SuperGPQA · Reason through hard problems	42.8	—	model author	2026-10-07
author-reported TAU1-Airline	32	—	model author	2026-10-07
author-reported TAU1-Retail	48.7	—	model author	2026-10-07
author-reported ZebraLogic	80.2	—	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	40.4	—	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	24	—	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	13.2	—	model author	2026-10-07

What the card says

2 cited sentences

Summary generated, cited

Qwen3-4B-Instruct-2507 is a 4.02B parameter causal language model designed for text generation.^[1]^[2] The model features a 262,144 native context length and is released under the Apache 2.0 license.^[3]^[4]

parameters human
“Type: Causal Language Models”
“262,144 natively”
license

Generated 2026-10-07 15:00 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

7 use cases

What it is good at 7 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.

Facts

21 fields

Listed in LLMs

Facts

Qwen3-4B-Instruct-2507 has 4.02B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2025-08-05, last modified 2025-09-17. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	Qwen3ForCausalLM
Model type	qwen3
Dtype	bfloat16
Params by dtype	BF16: 4,022,468,096
Size band	M
MoE	no
Context length	262,144 (config.json max_position_embeddings)
Papers	arXiv:2505.09388
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3ForCausalLM · 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 Qwen3-4B-Instruct-2507: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	9.7 GB · Single GPU (<24 GB)
FP8 (1 byte/param)	4.8 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)	2.4 GB · Edge (<4 GB)

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

All variants

40 variants

Lineage

Original model Family: Qwen/Qwen3 8,858 derivatives on the Hub

quantized: 352finetune: 2,618adapter: 5,853merge: 35

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
Qwen/Qwen3-4B-Instruct-2507-FP8	quantized	-FP8	4.41B	M	apache-2.0	Qwen	90	9M
MaziyarPanahi/Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	—	MaziyarPanahi	3	1.8M
unsloth/Qwen3-4B-Instruct-2507-unsloth-bnb-4bit	quantized	-bnb	4.02B	M	apache-2.0	unsloth	17	1.8M
unsloth/Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	apache-2.0	unsloth	220	1.5M
lmstudio-community/Qwen3-4B-Instruct-2507-MLX-4bit	quantized	MLX	4.02B	M	apache-2.0	lmstudio-community	3	1.2M
lmstudio-community/Qwen3-4B-Instruct-2507-MLX-8bit	quantized	MLX	4.02B	M	apache-2.0	lmstudio-community	2	1.2M
lmstudio-community/Qwen3-4B-Instruct-2507-MLX-6bit	quantized	MLX	4.02B	M	apache-2.0	lmstudio-community	0	1.2M
lmstudio-community/Qwen3-4B-Instruct-2507-MLX-5bit	quantized	MLX	4.02B	M	apache-2.0	lmstudio-community	0	1.1M
llamafactory/tiny-random-qwen3	finetune	—	2.44M	S	apache-2.0	llamafactory	0	954.3k
unsloth/Qwen3-4B-Instruct-2507	finetune	—	4.02B	M	apache-2.0	unsloth	26	738.4k
GeneralAnalysis/GA_Guard_Lite	finetune	—	596M	S	cc-by-nc-4.0	GeneralAnalysis	7	637.7k
allenai/Molmo2-4B	finetune	—	4.85B	M	apache-2.0	allenai	54	520.6k
GeneralAnalysis/GA_Guard_Core	finetune	—	4.02B	M	cc-by-nc-4.0	GeneralAnalysis	3	518.5k
warshanks/Qwen3-4B-Instruct-2507-AWQ	quantized	AWQ	4.47B	M	apache-2.0	warshanks	2	462.9k
cyankiwi/Qwen3-4B-Instruct-2507-AWQ-4bit	quantized	AWQ	4.41B	M	apache-2.0	cyankiwi	8	377.4k
mlx-community/Qwen3-4B-Instruct-2507-4bit	quantized	-4bit	4.02B	M	apache-2.0	mlx-community	15	360.3k
lmstudio-community/Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	—	lmstudio-community	19	351.1k
llmat/Qwen3-4B-Instruct-2507-NVFP4	quantized	NVFP4	2.82B	M	apache-2.0	llmat	2	294.4k
bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	—	bartowski	25	280.9k
JunHowie/Qwen3-4B-Instruct-2507-GPTQ-Int4	quantized	GPTQ	4.02B	M	apache-2.0	JunHowie	4	273.6k
unsloth/Qwen3-4B-Instruct-2507-bnb-4bit	quantized	-bnb	4.14B	M	apache-2.0	unsloth	5	234.9k
superjob/Qwen3-4B-Instruct-2507-GPTQ-Int4	quantized	GPTQ	4.44B	M	apache-2.0	superjob	1	101.6k
ystemsrx/Qwen3-Sex	quantized	GGUF	4.02B	M	apache-2.0	ystemsrx	26	78k
RefalMachine/RuadaptQwen3-4B-Instruct-GGUF	quantized	GGUF	—	—	apache-2.0	RefalMachine	10	67.5k
Manojb/Qwen3-4B-toolcalling-gguf-codex	quantized	gguf	—	—	mit	Manojb	56	56.1k
huihui-ai/Huihui-Qwen3-4B-Instruct-2507-abliterated	finetune	—	4.02B	M	apache-2.0	huihui-ai	45	54.6k
allenai/Molmo2-VideoPoint-4B	finetune	—	4.85B	M	apache-2.0	allenai	22	47.7k
Jackrong/GPT-5-Distill-Qwen3-4B-Instruct-GGUF	quantized	GGUF	—	—	apache-2.0	Jackrong	20	42.5k
HauhauCS/Qwen3-4B-2507-Instruct-Uncensored-HauhauCS-Aggressive	quantized	GGUF	—	—	apache-2.0	HauhauCS	27	40.8k
Eslzzyl/Qwen3-4B-Instruct-2507-AWQ	quantized	AWQ	4.02B	M	apache-2.0	Eslzzyl	1	36.1k
heretic-org/Qwen3-4B-Instruct-2507-heretic	finetune	—	4.02B	M	apache-2.0	heretic-org	21	34.9k
LocoreMind/LocoOperator-4B	finetune	—	4.02B	M	mit	LocoreMind	279	20.4k
byteshape/Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	apache-2.0	byteshape	39	26.3k
ggml-org/Qwen3-4B-Instruct-2507-Q8_0-GGUF	quantized	Q8_0	—	—	apache-2.0	ggml-org	0	25.3k
cyankiwi/Qwen3-4B-Instruct-2507-AWQ-8bit	quantized	AWQ	1.80B	S	apache-2.0	cyankiwi	1	18.9k
RefalMachine/RuadaptQwen3-4B-Instruct	finetune	—	4.01B	M	apache-2.0	RefalMachine	14	16k
AliesTaha/fable-traces	finetune	—	4.02B	M	apache-2.0	AliesTaha	209	6.2k
Alcoft/Qwen_Qwen3-4B-Instruct-2507-GGUF	quantized	GGUF	—	—	apache-2.0	Alcoft	0	15.6k
JunHowie/Qwen3-4B-Instruct-2507-GPTQ-Int8	quantized	GPTQ	4.02B	M	apache-2.0	JunHowie	1	14.8k
Manojb/Qwen3-4b-toolcall-gguf-llamacpp-codex	quantized	gguf	—	—	mit	Manojb	9	13.1k

Repository changelog

4 commits

changed in the last 7 daysNew providers: 2026-10-23Other: 2026-10-24all changes

Repository changelog 4 commits

Date	Commit	SHA
2025-09-17	Update tokenizer_config.json	cdbee75
2025-08-06	Update README.md	eb25fbe
2025-08-05	Upload folder using huggingface_hub	1b4199c
2025-08-05	initial commit	e7974da

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for Qwen3-4B-Instruct-2507, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.01/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Nscale hf	Qwen/Qwen3-4B-Instruct-2507	256K	—	$0.01	$0.03	23.5	386 ms	yes	live	2026-10-25

Articles and coverage

18 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.

Official1

Qwen3: Think deeper, act faster

qwenlm.github.io2025-04-28officialauthorityHN front pageabout the Qwen3 familyHN 869 pts · 388 comments2,139 words

Qwen3 introduces Hybrid Thinking Modes, allowing users to toggle between step-by-step reasoning and rapid responses via computational budget control. The open-weight MoE model Qwen3-235B-A22B achieves performance similar to dense models while using only 10% of the active parameters. Pre-training utilized approximately 36 trillion tokens across 119 languages, with deployment supported by SGLang and vLLM.written, reviewed, cited

Qwen3-8B and Qwen3-14B models do not use tied embeddings, unlike smaller variants.
The pre-training pipeline includes a final stage extending context length to 32K tokens.

Analysis9

Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers

venturebeat.com2026-02-28HN front pageabout the Qwen3 familyHN 461 pts · 272 comments

GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2

magazine.sebastianraschka.com2025-08-10authorityHN front pageabout the Qwen3 familyHN 490 pts · 97 comments5,195 words

OpenAI released gpt-oss-120b and gpt-oss-20b, their first open-weight models since GPT-2 in 2019. The 20B model can run on a consumer GPU with up to 16 GB of RAM, and the 120B model can run on a single H100 with 80 GB of RAM or newer hardware. The models utilize MXFP4 optimization to fit onto single GPUs.written, reviewed, cited

A 2025 research paper confirms Dropout results in worse downstream performance in single-epoch regimes.

Qwen3 4B 2507 Models - Intelligence, Performance & Price Comparison | Artificial Analysis

artificialanalysis.aiauthority

src/qai_hub_models/models/qwen3_4b_instruct_2507/README.md

github.comauthority

Qwen3 4B Instruct 2507: benchmarks, context window — ModelCap #160

modelcap.ai

Qwen3 4B 2507 Instruct | Benched.ai

benched.ai

Qwen : Qwen 3 4B Instruct 2507

ai-stats.phaseo.app

Qwen3-4B-Instruct-2507: Lightweight Open-Source LLM

dev.co

Qwen3-4B-Instruct-2507 — specs, VRAM, license, benchmarks | LocalLLMs

localllms.dev

All coverage tracked for this model (18) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Qwen3 4B 2507 Models - Intelligence, Performance & Price Comparison \| Artificial Analysis	artificialanalysis.ai	—	2026-10-07	exa
src/qai_hub_models/models/qwen3_4b_instruct_2507/README.md	github.com	—	2026-10-07	exa
Qwen3 4B Instruct 2507: benchmarks, context window — ModelCap #160	modelcap.ai	—	2026-10-07	exa
Qwen3 4B 2507 Instruct \| Benched.ai	benched.ai	—	2026-10-07	exa
Qwen : Qwen 3 4B Instruct 2507	ai-stats.phaseo.app	—	2026-10-07	exa
Qwen3-4B-Instruct-2507: Lightweight Open-Source LLM	dev.co	—	2026-10-07	exa
Qwen3-4B-Instruct-2507 — specs, VRAM, license, benchmarks \| LocalLLMs	localllms.dev	—	2026-10-07	exa
Qwen3 4B 2507 Instruct Review \| Pricing, Benchmarks & Capabilities (2026) \| Design for Online	designforonline.com	2026-04-24	2026-10-07	exa
Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers	venturebeat.com	2026-02-28	2026-10-07	hn
Qwen3-Coder-Next	qwen.ai official	2026-02-03	2026-10-07	hn
Qwen3-Max-Thinking	qwen.ai official	2026-01-26	2026-10-07	hn
Qwen3-TTS family is now open sourced: Voice design, clone, and generation	qwen.ai official	2026-01-22	2026-10-07	hn
Qwen3-VL	qwen.ai official	2025-09-23	2026-10-07	hn
Qwen3-Omni: Native Omni AI model for text, image and video	github.com	2025-09-22	2026-10-07	hn
Qwen3-Next	qwen.ai official	2025-09-12	2026-10-07	hn
GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2	magazine.sebastianraschka.com	2025-08-10	2026-10-07	hn
Qwen3-Coder: Agentic coding in the world	qwenlm.github.io official	2025-07-22	2026-10-07	hn
Qwen3: Think deeper, act faster	qwenlm.github.io official	2025-04-28	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).

Qwen3 Technical Report this model

An Yang, Anfeng Li, Baosong Yang +57 · 2025-05-14 · PDF · HF paper page (349 upvotes)

Similar models

3 models

Similar models

Phi-3-mini-4k-instruct · 3.82B
same size band (M), same task (text-generation), permissive license
Phi-3.5-mini-instruct · 3.82B
same size band (M), same task (text-generation), permissive license
Phi-4-mini-instruct · 3.84B
same size band (M), same task (text-generation), permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:32 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25

Follow Qwen3-4B-Instruct-2507

Watch this model

/feeds/models/Qwen--Qwen3-4B-Instruct-2507.xml(RSS 2.0) · every change to Qwen3-4B-Instruct-2507
/feeds/families/Qwen--Qwen3.xml(RSS 2.0) · every change across the Qwen3 family

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