Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › Qwen3 family

Qwen3-Omni-30B-A3B-Instruct

Hugging Face ↗ Follow changes

Qwen Apache 2.0 (apache-2.0) out 13 months ago

35.3B parameters, 3B active, several kinds of input and output.

Verified 2026-10-25 Rank #165 in LLMs · score 91.5 method v1 Sources 3 Downloads 30d 682k · #141 of 619 tracked

Can I run it?21GB at INT4Estimatefits a 24 GB cardmethodCommunity: qwen3-omni-30b-a3b-instruct — Parameters, Context, Benchmarks & Pricing | AI Atlas →
Can I use it?Apache 2.0 (apache-2.0)Permissivelicence file2026-10-24
Do people use it?682kdownloads, 30 days#141 of 619 we track · top 23%+10 likes, downloads −6% this week1,041 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.

18 best of 18 tracked

The 4 picks above, plus these 14. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. ArticleQwen3-Coder-Nextqwen.ai · 2026-02-03 · official
  2. ArticleQwen3-Max-Thinkingqwen.ai · 2026-01-26 · official
  3. ArticleQwen3-TTS family is now open sourced: Voice design, clone, and generationqwen.ai · 2026-01-22 · official
  4. ArticleQwen3: Think deeper, act fasterqwenlm.github.io · 2025-04-28 · 869 HN pts · official
  5. ArticleQwen3-Coder: Agentic coding in the worldqwenlm.github.io · 2025-07-22 · official
  6. ArticleQwen3-Nextqwen.ai · 2025-09-12 · official
  7. ArticleQwen3-VLqwen.ai · 2025-09-23 · official
  8. ArticleGPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2magazine.sebastianraschka.com · 2025-08-10 · 490 HN pts
  9. ArticleQwen3 Omni 30B A3B Instruct vs Qwen3 Omni 30B A3B Instruct: Model Comparisonartificialanalysis.ai · undated
  10. ArticleQwen3 Omni 30B A3B Instruct vs Qwen3 30B A3B 2507 Instruct: Model Comparisonartificialanalysis.ai · undated
  11. ArticleQwen3 Omni 30B A3B Instruct: benchmarks — ModelCap #100modelcap.ai · undated
  12. ArticleQwen3 Omni 30B A3B Benchmarks, Pricing & Specs — Vector Wirevectorwire.ai · undated
  13. ArticleQwen3-Omni-30B-A3B-Instruct Benchmark Scores & Performancebenchlm.ai · undated
  14. ArticleQwen3 Omni 30B A3B Instruct — Hardware Requirements & Compatibility | llmrunllmrun.dev · undated

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.

No API provider serves it, so its downloads are people running it themselves: 36 quantized builds and 28 Spaces.

  • As a service0no API provider listed
  • On own hardware36quantized builds
  • Built on6341 fine-tunes · 22 adapters · 0 merges
  • In apps28Hugging Face Spaces

Top Spaces running it: Qwen/Qwen3-Omni-Demo ♥ 276 · Mohaddz/Qwen3-Omni-Local-Demo ♥ 6 · chenxie95/Resonate ♥ 5 · BATUTO-ART/BATUTO_neurocoreWeb ♥ 1

Will it fit?

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

Fits a 64 GB Mac (unified memory) at INT4

About 21 GB at INT4, 42 GB at FP8, 85 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 21 GB of weights (estimate; KV cache and activations extra)INT421 GBFP8: about 42 GB of weights (estimate; KV cache and activations extra)FP842 GBBF16: about 85 GB of weights (estimate; KV cache and activations extra)BF1685 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 21 GB of weights (estimate; KV cache and activations extra)INT421 GBFP8: about 42 GB of weights (estimate; KV cache and activations extra)FP842 GBBF16: about 85 GB of weights (estimate; KV cache and activations extra)BF1685 GB

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

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 (2025-09-20), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.

MayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOct2025-09-22: 5 repo commits2025-09-21: 2 repo commits2025-09-20: 2 repo commitsQwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers · 2026-02-28Qwen3: Think deeper, act faster · 2025-04-28GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2 · 2025-08-10Qwen3-Omni: Native Omni AI model for text, image and video · 2025-09-22Weights 2025-09-20 · 2025-09-20Weights 2025-09-20now
MayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOct2025-09-22: 5 repo commits2025-09-21: 2 repo commits2025-09-20: 2 repo commitsQwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers · 2026-02-28Qwen3: Think deeper, act faster · 2025-04-28GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2 · 2025-08-10Qwen3-Omni: Native Omni AI model for text, image and video · 2025-09-22Weights 2025-09-20 · 2025-09-20Weights 2025-09-20now

What people built on it

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

99 derivatives on the Hub, led by 41 fine-tunes

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

36 quantized (Hugging Face model tree)41 fine-tunes (Hugging Face model tree)22 adapters (Hugging Face model tree)36 quantized41 fine-tunes22 adapters
36 quantized (Hugging Face model tree)41 fine-tunes (Hugging Face model tree)22 adapters (Hugging Face model tree)36 quantized41 fine-tunes22 adapters
GGUF ggml-org/Qwen3-Omni-30B-A3B-Instruct-GGUF for llama.cpp, Ollama, LM Studio 198.9k downloads MLX pherber3/Qwen3-Omni-30B-A3B-Instruct-4bit-mlx for Apple silicon 4.3k downloads AWQ cyankiwi/Qwen3-Omni-30B-A3B-Instruct-AWQ-4bit for vLLM, SGLang 545.7k downloads

All variants →

Reference

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

What the card says

1 cited sentence

Summary generated, cited

The model uses a mixture of experts architecture to process text, images, audio, and video.[1][2]

  1. is mixture of experts
  2. “It processes text, images, audio, and video”

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

3 use cases

What it is good at 3 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

19 fields

Listed in LLMs · MoE

Facts

Pipeline tagany-to-any
Librarytransformers
ArchitecturesQwen3OmniMoeForConditionalGeneration
Model typeqwen3_omni_moe
Dtypebfloat16
Params by dtypeBF16: 35,259,818,545
Size bandXL
MoEyes
Languagesen
LicenseApache 2.0 (apache-2.0) (other)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3OmniMoeForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3OmniMoeForConditionalGeneration · 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-Omni-30B-A3B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)85 GB · Multi-GPU node (<640 GB)
FP8 (1 byte/param)42 GB · Single high-end GPU (<80 GB)
INT4 (0.5 bytes/param)21 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original model Family: Qwen/Qwen3 99 derivatives on the Hub

quantized: 36finetune: 41adapter: 22

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
cyankiwi/Qwen3-Omni-30B-A3B-Instruct-AWQ-4bit quantized AWQ 10.0B L other cyankiwi 63 545.7k
marksverdhei/Qwen3-Omni-30B-A3B-FP8 quantized -FP8 35.3B XL other marksverdhei 6 209.2k
ggml-org/Qwen3-Omni-30B-A3B-Instruct-GGUF quantized GGUF — — — ggml-org 21 198.9k
cyankiwi/Qwen3-Omni-30B-A3B-Instruct-AWQ-8bit quantized AWQ 13.8B L other cyankiwi 8 85.5k
Atotti/Qwen3-Omni-AudioTransformer finetune — 648M S apache-2.0 Atotti 37 57.4k
Intel/Qwen3-Omni-30B-A3B-Instruct-int4-AutoRound quantized int4 9.36B L apache-2.0 Intel 5 42.1k
YihongJin/Qwen3-Omni-30B-A3B-Instruct-NVFP4-W4A4-full-thinker-awqclip quantized NVFP4 20.3B L apache-2.0 YihongJin 1 37.6k
shunyang90/Qwen3-Omni-30B-A3B-Instruct-NVFP4 finetune NVFP4 21.7B L apache-2.0 shunyang90 1 23.6k
TrevorJS/Qwen3-Omni-30B-A3B-GGUF quantized GGUF — — — TrevorJS 13 20.3k
Cloudflare/clef-omni finetune — 35.3B XL apache-2.0 Cloudflare 139 6.4k
catplusplus/Qwen3-Omni-30B-A3B-Instruct-NVFP4 quantized NVFP4 35.3B XL apache-2.0 catplusplus 0 5.4k
pherber3/Qwen3-Omni-30B-A3B-Instruct-4bit-mlx quantized -4bit 30.5B L apache-2.0 pherber3 8 4.3k
catplusplus/Qwen3-Omni-30B-A3B-Instruct-NVFP4-talker-safe quantized NVFP4 35.3B XL apache-2.0 catplusplus 0 4.3k
YihongJin/Qwen3-Omni-30B-A3B-Instruct-NVFP4-W4A16-awq quantized NVFP4 20.3B L apache-2.0 YihongJin 1 3.8k
rodial/Qwen3-Omni-30B-A3B-Thinking-GGUF quantized GGUF — — apache-2.0 rodial 1 3.8k
abnormalmapstudio/Qwen3-Omni-30B-A3B-Instruct-mxfp4-mlx quantized mxfp4 30.5B L other abnormalmapstudio 2 3.6k
Saktsant/Qwen3-Omni-30B-A3B-Instruct-AWQ quantized AWQ 10.0B L other Saktsant 2 3.3k
88plug/Qwen3-Omni-30B-W4A16 quantized -W4A16 35.3B XL apache-2.0 88plug 2 3.3k
steampunque/Qwen3-Omni-30B-A3B-Instruct-MP-GGUF quantized GGUF — — apache-2.0 steampunque 1 3.1k
NandemoGHS/Anime-Speech-Japanese-Refiner finetune — 35.3B XL cc-by-nc-4.0 NandemoGHS 13 2.4k
mohhaamed/Qwen3-Omni-30B-A3B-FP8 quantized -FP8 35.3B XL other mohhaamed 2 2.5k
DeepMentor/qwen3-omni-30B-a3b-instruct-gguf quantized gguf — — other DeepMentor 2 2.2k
jart25/Qwen3-Omni-30B-A3B-Instruct-AWQ-W4A16 quantized AWQ 10.0B L — jart25 1 2k
88plug/Qwen3-Omni-30B-A3B-W8A16 quantized — 34.6B L apache-2.0 88plug 0 1.1k
CMKL/MANGO-Qwen3-Omni-30B-A3B-Instruct finetune — 35.3B XL other CMKL 0 1k
cudabenchmarktest/Qwen3.8-27B-E03-Obliterated-Omni-GGUF quantized GGUF — — apache-2.0 cudabenchmarktest 2 919
MIThridate/Qwen3-Omni-35B-A3B-SAMI-AsymQ quantized GGUF — — apache-2.0 MIThridate 1 611
cudabenchmarktest/Qwen3.8-27B-E03-Obliterated-Omni-Audio-Bridge-GGUF quantized GGUF — — other cudabenchmarktest 0 659
zenlm/zen-omni finetune — 35.3B XL apache-2.0 zenlm 0 649
YihongJin/Qwen3-Omni-30B-A3B-Instruct-NVFP4-W4A4-experts-mse quantized NVFP4 20.8B L apache-2.0 YihongJin 0 611
cudabenchmarktest/Ornith-1.5-9B-Omni-GGUF quantized GGUF — — other cudabenchmarktest 0 609
YihongJin/Qwen3-Omni-30B-A3B-Instruct-NVFP4-W4A16-max quantized NVFP4 20.3B L apache-2.0 YihongJin 1 546
cy0307/ropedia-qwen3-omni-lora-128ep adapter — — — other cy0307 1 532
deepcrayon/AniMUL-v1 finetune — 35.3B XL apache-2.0 deepcrayon 6 251
ASLP-lab/FM-Speech finetune — 35.3B XL apache-2.0 ASLP-lab 6 175
Rakancorle11/qwen3omni_full_sft_with_audio finetune — 1.21M S other Rakancorle11 0 448
zenlm/zen-omni-30b-instruct finetune — 35.3B XL apache-2.0 zenlm 0 429
zenlm/zen3-omni finetune — 35.3B XL apache-2.0 zenlm 1 338
lainlives/Qwen3-Omni-30B-A3B-Instruct-bnb-4bit quantized -bnb 35.3B XL other lainlives 0 368
meetween/SpeechLMM-v2.0-XL-30B finetune — 35.4B XL cc-by-nc-sa-4.0 meetween 0 365

Repository changelog

9 commits

Repository changelog 9 commits

DateCommitSHA
2025-09-22 Update README.md 26291f7
2025-09-22 Update README.md 247db2c
2025-09-22 Update README.md 2c2d1c9
2025-09-22 Update README.md 357d06a
2025-09-22 Update config.json 12a3ac6
2025-09-21 Update README.md 06e43d9
2025-09-21 Update README.md a2de1f0
2025-09-20 Upload folder using huggingface_hub 0e20e2b
2025-09-20 initial commit 302ffa9

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-Omni: Native Omni AI model for text, image and video
github.com2025-09-22authorityHN front pageabout the Qwen3 familyHN 571 pts · 142 comments6,753 words
Qwen3 Omni 30B A3B Instruct vs Qwen3 Omni 30B A3B Instruct: Model Comparison
artificialanalysis.aiauthority
Qwen3 Omni 30B A3B Instruct vs Qwen3 30B A3B 2507 Instruct: Model Comparison
artificialanalysis.aiauthority
qwen3-omni-30b-a3b-instruct — Parameters, Context, Benchmarks & Pricing | AI Atlas
ai-atlas.co
Qwen3 Omni 30B A3B Instruct: benchmarks — ModelCap #100
modelcap.ai
Qwen3 Omni 30B A3B Benchmarks, Pricing & Specs — Vector Wire
vectorwire.aiabout the Qwen3 family
Qwen3-Omni-30B-A3B-Instruct — benchmarks, price and context
modelscale.dev
All coverage tracked for this model (18) — append-only, nothing is ever removed
TitleSourcePublishedFirst seenSignals
Qwen3 Omni 30B A3B Instruct vs Qwen3 Omni 30B A3B Instruct: Model Comparisonartificialanalysis.ai—2026-10-07exa
Qwen3 Omni 30B A3B Instruct vs Qwen3 30B A3B 2507 Instruct: Model Comparisonartificialanalysis.ai—2026-10-07exa
qwen3-omni-30b-a3b-instruct — Parameters, Context, Benchmarks & Pricing | AI Atlasai-atlas.co—2026-10-07exa
Qwen3 Omni 30B A3B Instruct: benchmarks — ModelCap #100modelcap.ai—2026-10-07exa
Qwen3 Omni 30B A3B Benchmarks, Pricing & Specs — Vector Wirevectorwire.ai—2026-10-07exa
Qwen3-Omni-30B-A3B-Instruct — benchmarks, price and contextmodelscale.dev—2026-10-07exa
Qwen3-Omni-30B-A3B-Instruct Benchmark Scores & Performancebenchlm.ai—2026-10-07exa
Qwen3 Omni 30B A3B Instruct — Hardware Requirements & Compatibility | llmrunllmrun.dev—2026-10-07exa
Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computersventurebeat.com2026-02-282026-10-07hn
Qwen3-Coder-Nextqwen.ai official2026-02-032026-10-07hn
Qwen3-Max-Thinkingqwen.ai official2026-01-262026-10-07hn
Qwen3-TTS family is now open sourced: Voice design, clone, and generationqwen.ai official2026-01-222026-10-07hn
Qwen3-VLqwen.ai official2025-09-232026-10-07hn
Qwen3-Omni: Native Omni AI model for text, image and videogithub.com2025-09-222026-10-07hn
Qwen3-Nextqwen.ai official2025-09-122026-10-07hn
GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2magazine.sebastianraschka.com2025-08-102026-10-07hn
Qwen3-Coder: Agentic coding in the worldqwenlm.github.io official2025-07-222026-10-07hn
Qwen3: Think deeper, act fasterqwenlm.github.io official2025-04-282026-10-07hn

Similar models

3 models

Similar models

  • Qwen3.6-35B-A3B · 36.0B
    same size band (XL), also MoE, more providers, permissive license
  • Qwen3.5-35B-A3B · 36.0B
    same size band (XL), also MoE, more providers, permissive license
  • GLM-4.7-Flash · 31.2B
    also MoE, more providers, permissive license

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 Qwen3-Omni-30B-A3B-Instruct

Watch this model

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