alt.hn

12/10/2025 at 4:13:38 PM

Qwen3-Omni-Flash-2025-12-01：a next-generation native multimodal large model

 https://qwen.ai/blog?id=qwen3-omni-flash-20251201

by pretext

12/10/2025 at 5:37:33 PM

This is a 30B parameter MoE with 3B active parameters and is the successor to their previous 7B omni model. [1]

You can expect this model to have similar performance to the non-omni version. [2]

There aren't many open-weights omni models so I consider this a big deal. I would use this model to replace the keyboard and monitor in an application while doing the heavy lifting with other tech behind the scenes. There is also a reasoning version, which might be a bit amusing in an interactive voice chat if it pronounces the thinking tokens while working through to a final answer.

1. https://huggingface.co/Qwen/Qwen2.5-Omni-7B

2. https://artificialanalysis.ai/models/qwen3-30b-a3b-instruct

by gardnr

12/10/2025 at 7:35:51 PM

This is a stack of models:

- 650M Audio Encoder

- 540M Vision Encoder

- 30B-A3B LLM

- 3B-A0.3B Audio LLM

- 80M Transformer/200M ConvNet audio token to waveform

This is a closed source weight update to their Qwen3-Omni model. They had a previous open weight release Qwen/Qwen3-Omni-30B-A3B-Instruct and a closed version Qwen3-Omni-Flash.

You basically can't use this model right now since none of the open source inference framework have the model fully implemented. It works on transformers but it's extremely slow.

by red2awn

12/10/2025 at 6:05:06 PM

Looks like it's not open source: https://www.alibabacloud.com/help/en/model-studio/qwen-omni#...

by olafura

12/10/2025 at 6:11:30 PM

No... that website is not helpful. If you take it at face value, it is claiming that the previous Qwen3-Omni-Flash wasn't open either, but that seems wrong? It is very common for these blog posts to get published before the model weights are uploaded.

by coder543

12/10/2025 at 7:37:25 PM

The previous -Flash weight is closed source. They do have weights for the original model that is slightly behind in performance https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct

by red2awn

12/10/2025 at 8:00:19 PM

Based on things I had read over the past several months, Qwen3-Flash seemed to just be a weird marketing term for the Qwen3-Omni-30B-A3B series, not a different model. If they are not the same, then that is interesting/confusing.

by coder543

12/10/2025 at 8:15:34 PM

It is an in-house closed weight model for their own chat platform, mentioned in Section 5 of the original paper: https://arxiv.org/pdf/2509.17765

I've seen it in their online materials too but can't seem to find it now.

by red2awn

12/10/2025 at 5:56:49 PM

I can't find the weights for this new version anywhere. I checked modelscope and huggingface. It looks like they may have extended the context window to 200K+ tokens but I can't find the actual weights.

by gardnr

12/10/2025 at 6:22:11 PM

> There is also a reasoning version, which might be a bit amusing in an interactive voice chat if it pronounces the thinking tokens while working through to a final answer.

last i checked (months ago) claude used to do this

by tensegrist

12/10/2025 at 7:40:06 PM

I dont think the Flash model discussed in the article is 30B

Their benchmark table shows it beating Qwen3-235B-A22B

Does "Flash" in the name of a Qwen model indicate a model-as-a-service and not open weights?

by plipt

12/10/2025 at 7:49:36 PM

Flash is a closed weight version of https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct (it is 30B but with addtional training on top of the open weight release). They deploy the flash version on Qwen's own chat.

by red2awn

12/10/2025 at 7:59:48 PM

Thanks

Was it being closed weight obvious to you from the article? Trying to understand why I was confused. Had not seen the "Flash" designation before

Also 30B models can beat a semi-recent 235B with just some additional training?

by plipt

12/10/2025 at 8:19:56 PM

They had a Flash variant released alongside the original open weight release. It is also mentioned in Section 5 of the paper: https://arxiv.org/pdf/2509.17765

For the evals it's probably just trained on a lot of the benchmark adjacent datasets compared to the 235B model. Similar thing happened on other model today: https://x.com/NousResearch/status/1998536543565127968 (a 30B model trained specifically to do well in maths get near SOTA scores)

by red2awn

12/11/2025 at 9:54:44 AM

Haha, you could hear how it’s mind thinks, maybe by putting a lot of reverb on the thinking tokens or some other effect…

by andy_ppp

12/10/2025 at 6:37:01 PM

> This is a 30B parameter MoE with 3B active parameters

Where are you finding that info? Not saying you're wrong; just saying that I didn't see that specified anywhere in the linked page, or on their HF.

by andy_xor_andrew

12/13/2025 at 5:20:33 PM

I was wrong. I confused this with their open model. Looking at it more closely, it is likely an omni version of Qwen3-235B-A22B. I wonder why they benchmarked it against Qwen2.5-Omni-7B instead of Qwen3-Omni-30B-A3B.

I wish I could delete the comment.

by gardnr

12/10/2025 at 7:36:17 PM

The link[1] at the top of their article to HuggingFace goes to some models named Qwen3-Omni-30B-A3B that were last updated in September. None of them have "Flash" in the name.

The benchmark table shows this Flash model beating their Qwen3-235B-A22B. I dont see how that is possible if it is a 30B-A3B model.

I don't see a mention of a parameter count anywhere in the article. Do you? This may not be an open weights model.

This article feels a bit deceptive

1: https://huggingface.co/collections/Qwen/qwen3-omni

by plipt

12/10/2025 at 4:55:48 PM

Does Qwen3-Omni support real-time conversation like GPT-4o? Looking at their documentation it doesn't seem like it does.

Are there any open weight models that do? Not talking about speech to text -> LLM -> text to speech btw I mean a real voice <-> language model.

edit:

It does support real-time conversation! Has anybody here gotten that to work on local hardware? I'm particularly curious if anybody has run it with a non-nvidia setup.

by sosodev

12/10/2025 at 10:34:37 PM

From what I can tell, their official chat site doesn't have a native audio -> audio model yet. I like to test this through homophones (e.g. record and record) and asking it to change its pitch or produce sounds.

by potatoman22

12/11/2025 at 3:59:16 AM

“record and record”, if you mean the verb for persisting something and the noun for the thing persisted, are heteronyms (homographs which are not homophones), which incidentally is also what you would probably want to test what you are talking about here (distinguishing homophones would test use of context to understand meaning, but wouldn’t test anything about whether or not logic was working directly on audio or only working on text processed from audio, failing to distinguish heteronyms is suggestive of processing occurring on text, not audio directly.)

by dragonwriter

12/11/2025 at 4:21:00 AM

There are homophones of “record”, such as:

“He’s on record saying he broke the record for spinning a record.”

by bakeman

12/11/2025 at 5:13:01 AM

True.

OTOH my point that the thing being suggested to be tested is not testable by seeing whether or not the system is capable of distinguishing homophones, but might be by seeing whether or not it distingishes heteronyms still stands. (The speculation that the record/record distinction intended was one that is actually a pair of heteronyms and that the error was merely the use of the word “homophone" in place of “heteronym”, rather than the basic logic of the comment is somewhat tangential to the main point.)

by dragonwriter

12/11/2025 at 10:37:48 PM

Ah I meant heteronyms. Thanks!

by potatoman22

12/10/2025 at 11:58:05 PM

Huh, you're right. I tried your test and it clearly can't understand the difference between homophones. That seems to imply they're using some sort of TTS mechanism. Which is really weird because Qwen3-Omni claims to support direct audio input into the model. Maybe it's a cost saving measure?

by sosodev

12/11/2025 at 5:19:41 AM

Weirdly, I just tried it again and it seems to understand the difference between record and record just fine. Perhaps if there's heavy demand for voice chat, like after a new release, they load shed by using TTS to a smaller model.

However, It still doesn't seem capable of producing any of the sounds, like laughter, that I would expect from a native voice model.

by sosodev

12/11/2025 at 10:38:50 PM

To be fair, discerning heteronyms might just be a gap in its training.

by potatoman22

12/11/2025 at 3:53:37 AM

Is record a homophone? At least in the UK we use different pronunciations for the meanings. Re-cord for the verb, rec-ord for the noun.

by djtango

12/11/2025 at 10:36:18 PM

I was mistaken about what homophone means!

by potatoman22

12/10/2025 at 7:38:25 PM

None of inference frameworks (vLLM/SGLang) supports the full model, let alone non-nvidia.

by red2awn

12/10/2025 at 8:50:21 PM

We actually deployed working speech to speech inference that builds on top of vLLM as the backbone. The main thing was to support the "Talker" module, which is currently not supported on the qwen3-omni branch for vLLM.

Check it out here: https://models.hathora.dev/model/qwen3-omni

by AndreSlavescu

12/10/2025 at 9:49:08 PM

Is your work open source?

by sosodev

12/10/2025 at 9:08:52 PM

Nice work. Are you working on streaming input/output?

by red2awn

12/10/2025 at 9:14:31 PM

Yeah, that's something we currently support. Feel free to try the platform out! No cost to you for now, you just need a valid email to sign up on the platform.

by AndreSlavescu

12/11/2025 at 10:07:57 AM

I tried this out, and it's not passing the record (n.) vs. record (v.) test mentioned elsewhere in this thread. (I can ask it to repeat one, and it often repeats the other.) Am I not enabling the speech-to-speech-ness somehow?

by valleyer

12/10/2025 at 8:29:12 PM

That's unfortunate but not too surprising. This type of model is very new to the local hosting space.

by sosodev

12/11/2025 at 2:23:42 AM

Makes sense, I think streaming audio->audio inference is a relatively big lift.

by whimsicalism

12/11/2025 at 9:21:18 AM

Correct, it's breaks the single prompt, single completion assumption baked into the frameworks. Conceptually it's still prompt/completion but for low latency response you have to do streaming KV cache prefill with a websocket server.

by red2awn

12/11/2025 at 5:37:21 PM

I imagine you have to start decoding many speculative completions in parallel to have true low latency.

by whimsicalism

12/10/2025 at 8:16:44 PM

That's exciting. I doubt there are any polished voice chat local apps yet that you can easily plug this into (I doubt the user experience is "there" yet). Even stuff like Silly Tavern is near unusable, lots of work to be done on the local front. Local voice models are what's going to enable that whole Minority Report workflow soon enough (especially if commands and intent are determined at the local level, and the meat of the prompt is handled by a larger remote model).

This is part of programming that I think is the new field. There will be tons of work for those that can build the new workflows which will need to be primarily natural language driven.

by ivape

12/10/2025 at 5:01:25 PM

it seems to be able to do native speech-speech

by dsrtslnd23

12/10/2025 at 5:12:19 PM

It does for sure. I did some more digging and it does real-time too. That's fascinating.

by sosodev

12/10/2025 at 5:54:43 PM

Is there a way to run these Omni models on a Macbook quantized via GGUF or MLX? I know I can run it in LMStudio or Llama.cpp but they don't have streaming microphone support or streaming webcam support.

Qwen usually provides example code in Python that requires Cuda and a non-quantized model. I wonder if there is by now a good open source project to support this use case?

by terhechte

12/10/2025 at 7:39:06 PM

Yes - there is a way: https://github.com/ggml-org/whisper.cpp

by mobilio

12/10/2025 at 8:21:24 PM

Whisper and Qwen Omni models have completely different architectures as far as I know

by novaray

12/10/2025 at 5:51:43 PM

The main issue I'm facing with realtime responses (speech output) is how to separate non-diegetic outputs (e.g thinking, structured outputs) from outputs meant to be heard by the end user.

I'm curious how anyone has solved this

by sim04ful

12/10/2025 at 7:19:17 PM

A simple way is to split the model’s output stream before TTS. Reasoning/structured tokens go into one bucket, actual user-facing text into another. Only the second bucket is synthesized. Most thinking out loud issues come from feeding the whole stream directly into audio.

by artur44

12/10/2025 at 8:19:13 PM

There is no TTS here. It's a native audio output model which outputs audio tokens directly. (At least, that's how the other real-time models work. Maybe I've misunderstood the Qwen-Omni architecture.)

by pugio

12/10/2025 at 8:35:01 PM

True, but even with native audio-token models you still need to split the model’s output channels. Reasoning/internal tokens shouldn't go into the audio stream only user-facing content should be emitted as audio. The principle is the same, whether the last step is TTS or audio token generation.

by artur44

12/11/2025 at 3:32:19 PM

There's an assumption there that the audio stream contains an equivalent of the <think>/</think> tokens. Every reason to think it should, but without seeing the tokeniser config it's a bit of a guess.

by regularfry

12/10/2025 at 6:57:55 PM

Wow, just 32B? This could almost run on a good device with 64 GB RAM. Once it gets to Ollama I'll have to see just what I can get out of this.

by devinprater

12/10/2025 at 7:23:30 PM

I see that their HuggingFace link goes to some Qwen3-Omni-30B-A3B models that show a last updated date of September

The benchmark table in their article shows Qwen3-Omni-Flash-2025-12-01 (and the previous Flash) as beating Qwen3-235B-A22B. How is that possible if this is only a 30B-A3B model? Also confusing how that comparison column starts out with one model but changes them as you descend down the table.

I don't see any FLASH variant listed on their Hugginface. Am i just missing it or are these specifying a model only used for their API service and there are no open weights to download?

by plipt

12/10/2025 at 7:28:59 PM

I run these on a 48gb Mac because of the universal ram.

by apexalpha

12/10/2025 at 6:03:04 PM

Looks to be API only. Bummer.

by aschobel

12/10/2025 at 8:05:20 PM

The models are right here, one of the first links in the post: https://huggingface.co/collections/Qwen/qwen3-omni

edit: Nevermind, in spite of them linking it at the top, they are the old models. Also, the HF demo is calling their API and not using HF for compute.

by readyplayeremma

12/10/2025 at 10:00:44 PM

It is super confusing. I also thought this initially was open weights.

by aschobel

12/11/2025 at 8:15:56 AM

Having lots of success with Gemini Flash Live 2.5. I am hoping 3.0 to come out soon. Benchmarks here claim better results that Gemini Live but have to test it. In past I've always been disappointed with Qwen Omni models in my English-first case...

by mohsen1

12/10/2025 at 4:58:28 PM

Does anyone else find that there's hard to pin down reason of life-lessness in the speech of these voice models?

Especially in the fruit pricing portion of the video for this model. Sounds completely normal but I can immediately tell it is ai. Maybe it's intonation or the overly stable rate of speech?

by binsquare

12/10/2025 at 5:10:43 PM

IMHO it's not lifeless. It's just not overly emotional. I definitely prefer it that way. I do not want the AI to be excited. It feels so contrived.

On the video itself: Interesting, but "ideal" was pronounced wrong in German. For a promotional video, they should have checked that with native speakers. On the other hand its at least honest.

by Lapel2742

12/10/2025 at 6:29:34 PM

I hate with a passion the over-americanized "accent" of chatgpt voices. Give me a bland one any day of the week

by nunodonato

12/11/2025 at 7:31:57 AM

Yeah that overly fake-excited voice type. Doesn't work for Europe at all. But indeed common in American customer service scenarios.

by wkat4242

12/10/2025 at 5:09:11 PM

I think it's because they've crammed vision, audio, multiple voices, prosody control, multiple languages, etc into just 30 billion parameters.

I think ChatGPT has the most lifelike speech with their voice models. They seem to have invested heavily in that area while other labs focused elsewhere.

by sosodev

12/10/2025 at 7:54:11 PM

I'm not convinced its end-to-end multimodal - in that case, you'll have a speech synthesis section and this will be some of the result. You could test by having it sing or do some accents, or have it talk back to you in an accent you give it.

by vessenes

12/10/2025 at 5:06:07 PM

> Sounds completely normal but I can immediately tell it is ai.

Maybe that's a good thing?

by esafak

12/10/2025 at 5:04:24 PM

I'm perfectly ok with and would prefer an AI "accent".

by colechristensen

12/10/2025 at 5:20:27 PM

Wow, crushing 2.5 Flash on every benchmark is huge. Time to move all of my LLM workloads to a local GPU rig.

by banjoe

12/10/2025 at 5:27:50 PM

Just remember to benchmark it yourself first with you private task collection, so you can actually measure them against each other. Pretty much any public benchmark is unreliable at this moment, and making model choices based on other's benchmarks is bound to leave you disappointed.

by embedding-shape

12/10/2025 at 7:10:58 PM

This. Last benchmarks of DSv3.2spe hinted at beating basically everything, yet in my testing even sonnet is miles ahead both in terms of speed and accuracy

by MaxikCZ

12/10/2025 at 8:45:54 PM

Why would you use an Omni model for text only workload... There is Qwen3-30B-A3B.

by red2awn

12/11/2025 at 10:41:36 AM

Except the image benchmarks are compared against 2.0, which seems suspicious that they would casually drop to an older model for those.

by skrunch

12/11/2025 at 1:34:30 AM

I truly enjoy how the naming conventions seem to follow how I did homework assignments back in the day: finalpaper-1-dec2nd, finalpaper-2-dec4th, etc etc.

by forgingahead

12/10/2025 at 4:45:29 PM

I asked: "How many resistors are used in fuzzhugger phantom octave guitar pedal?". It replied 29 resistors and provided a long list. Answer is 2 resistors: https://tagboardeffects.blogspot.com/2013/04/fuzzhugger-phan...

by dvh

12/10/2025 at 4:48:32 PM

> How many resistors are used in fuzzhugger phantom octave guitar pedal?

Weird, as someone not having a database of the web, I wouldn't be able to calculate either result.

by iFire

12/10/2025 at 4:57:04 PM

"I don't know" would be perfectly reasonable answer

by dvh

12/10/2025 at 7:08:49 PM

I feel like theres a time in near future where LLMs will be too cautious to answer any questions they arent sure about, and most of the human effort will go into pleading the LLM to at least try to give an answer, which will almost always be correct anyways.

by MaxikCZ

12/10/2025 at 8:44:08 PM

That would be a great if you could have a setting like temperature 0.0-1.0 (Only answer if you are 100% to guess as much as you like).

by plufz

12/10/2025 at 9:50:04 PM

It's not going to happen as the user would just leave the platform.

It would be better for most API usage though, as for business doing just a fraction of the job with 100% accuracy is often much preferable than claiming to do 100% but 20% is garbage.

by littlestymaar

12/10/2025 at 5:36:21 PM

> as someone not having a database of the web, I wouldn't be able to calculate either result

And that's how I know you're not an LLM!

by kaoD

12/10/2025 at 4:49:09 PM

I tend to pick things where I think the answer is in the introduction material like exams that test what was taught.

by iFire

12/10/2025 at 4:56:29 PM

This is just trivia. I would not use it to test computers -- or humans.

by esafak

12/10/2025 at 6:58:24 PM

It's good way to assess the model with respect to hallucinations though.

I don't think a model should know the answer, but it must be able to know that it doesn't know if you want to use it reliably.

by littlestymaar

12/10/2025 at 7:07:22 PM

No model is good at this yet. I'd expect the flagships to solve the first.

by esafak

12/10/2025 at 5:09:14 PM

Everything is just trivia until you have a use for the answer.

OP provided a we link with the answer, aren't these models supposed to be trained on all of that data?

by parineum

12/10/2025 at 5:16:31 PM

There is nothing useful you can do with this information. You might as well memorize the phone book.

The model has a certain capacity -- quite limited in this case -- so there is an opportunity cost in learning one thing over another. That's why it is important to train on quality data; things you can build on top of.

by esafak

12/11/2025 at 2:14:34 AM

What if you are trying to fix one of these things and needed a list of replacement parts?

by parineum

12/11/2025 at 2:28:08 AM

Not the right problem for this model. Any RAG-backed SLM would do; the important part is being backed by a search engine, like https://google.com/ai

by esafak

12/10/2025 at 6:03:40 PM

Just because it's in the training data doesn't mean the model can remember it. The parameters total 60 gigabytes, there's only so much trivia that can fit in there so it has to do lossy compression.

by DennisP

12/10/2025 at 4:58:50 PM

Where did you try it? I don’t see this model listed in the linked Qwen chat.

by brookst

12/11/2025 at 2:19:15 AM

Lol I asked it how many rooms I have in my house and it got that wrong. Llms are useless amirite

by bongodongobob

12/11/2025 at 2:22:50 AM

[dead]

by cindyllm

12/10/2025 at 7:30:35 PM

Maybe it thinks some of those 29 are in series:)

by strangattractor

12/10/2025 at 4:49:10 PM

I wonder if with that music analysis mode, you can also make your own synths

by mettamage

12/10/2025 at 5:43:48 PM

Is this a new proprietary model?

by Aissen

12/10/2025 at 5:01:42 PM

GPT4o in the charts is crazy.

by rarisma

12/10/2025 at 5:11:32 PM

Why? gpt-realtime is finalized gpt-4o. Gemini Live is still 2.5.

Not their fault frontier labs are letting their speech to speech offerings languish.

by BoorishBears

12/11/2025 at 2:20:06 PM

Qwen seem to be deliberately confusing about if they are releasing models open weight or not. I think largely not any more and you can go on quite a wild goose chase looking for different things that are implied they are released but are actually only available via API.

by andy_ppp

12/10/2025 at 7:52:29 PM

Interesting - when I asked the omni model at qwen.com what version it was, I got a testy "I don't have a version" and then was told my chat was blocked for inappropriate content. A second try asking for knowledge cutoff got me the more equivocal "2024, but I know stuff after that date, too".

No idea how to check if this is actually deployed on qwen.com right now.

by vessenes

12/10/2025 at 8:01:18 PM

> No idea how to check if this is actually deployed on qwen.com right now.

Assuming you mean qwen.ai, when you run a query it should take you to chat.qwen.ai with the list of models in the top left. None of the options appear to be the -Omni variant (at least when anonymously accessing it).

by zamadatix

12/10/2025 at 8:03:11 PM

Thanks - yes - I did. The blog post suggests clicking the 'voice' icon on the bottom right - that's what I did.

by vessenes

12/10/2025 at 9:49:46 PM

For what it's worth, that's not a reliable way to check what model you're interacting with.

by mh-

12/11/2025 at 6:08:34 PM

It’s a good positive signal, but not a good negative one.

It would be convincing if it said “I’m qwen-2025-12-whatever”. I agree it’s not dispositive if it refuses or claims to be llama 3 say. Generally most models I talk to do not hallucinate future versions of themselves, in fact it can be quite difficult to get them to use recent model designations; they will often autocorrect to older models silently.

by vessenes