Up to date
Qwen Qwen License (qwen-research) out 18 months ago
5.54B parameters, several kinds of input and output.
Verified 2026-10-25 Rank #248 in LLMs · score 86.1 method v1 Sources 3 Downloads 30d 144.9k · #274 of 619 tracked
Picked by rule from 32 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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: 15 quantized builds and 26 Spaces.
Top Spaces running it: tencent/AuK ♥ 127 · FunAudioLLM/PrismAudio ♥ 36 · hugging-apps/drbaph-auk-comfyui ♥ 4 · jawakja/Mulktimodal_chatbot ♥ 2
Weights only, an estimate: leave room for the KV cache. Method
About 3.3 GB at INT4, 6.6 GB at FP8, 13 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2025-04-30), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
92 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported OmniBench | 52.14%|52.08%|52.83%|52.19% percent | Speech | Sound Event | Music | Avg | model author | 2026-10-07 |
The model is designed to perceive text, images, audio, and video while generating text and natural speech responses.[1]
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.
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.
Listed in LLMs
| Pipeline tag | any-to-any |
|---|---|
| Library | transformers |
| Architectures | Qwen2_5OmniModel |
| Model type | qwen2_5_omni |
| Dtype | bfloat16 |
| Params by dtype | BF16: 5,088,069,376 · F32: 449,051,296 |
| Size band | M |
| MoE | no |
| Languages | en |
| Papers | arXiv:2503.20215 |
| License | Qwen License (qwen-research) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2_5OmniModel · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: not in registry · 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.
Weights-only memory for Qwen2.5-Omni-3B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 13 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 6.6 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 3.3 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 15finetune: 37adapter: 39merge: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| ggml-org/Qwen2.5-Omni-3B-GGUF | quantized | GGUF | — | — | other | ggml-org | 16 | 122.1k |
| unsloth/Qwen2.5-Omni-3B-GGUF | quantized | GGUF | — | — | other | unsloth | 57 | 87.5k |
| mlboydaisuke/Qwen2.5-Omni-3B-Audio-CoreAI | quantized | — | — | — | other | mlboydaisuke | 2 | 17.4k |
| mradermacher/Qwen2.5-Omni-3B-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 3 | 9.1k |
| mradermacher/Qwen2.5-Omni-3B-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 7.1k |
| AudioVisual-Caption/ASID-Captioner-3B | finetune | — | 4.70B | M | apache-2.0 | AudioVisual-Caption | 37 | 3.9k |
| openinterx/UGC-VideoCaptioner | finetune | — | 5.54B | M | mit | openinterx | 5 | 4k |
| Haon-Chen/e5-omni-3B | finetune | — | 4.70B | M | mit | Haon-Chen | 9 | 3.5k |
| admiralakber/Qwen2.5-Omni-3B-Q8_0-GGUF | quantized | Q8_0 | — | — | other | admiralakber | 0 | 2.2k |
| KE-Team/Ke-Omni-R-3B | finetune | — | 4.70B | M | apache-2.0 | KE-Team | 23 | 757 |
| unsloth/Qwen2.5-Omni-3B | finetune | — | 5.54B | M | other | unsloth | 5 | 1.6k |
| steampunque/Qwen2.5-Omni-3B-MP-GGUF | quantized | GGUF | — | — | apache-2.0 | steampunque | 1 | 1.5k |
| giangndm/qwen2.5-omni-3b-mlx-8bit | finetune | mlx | 4.03B | M | other | giangndm | 2 | 1.4k |
| yuhong123/Qwen2.5-Omni-3B-Q4_K_M-GGUF | quantized | GGUF | — | — | other | yuhong123 | 0 | 962 |
| giangndm/qwen2.5-omni-3b-mlx-4bit | finetune | mlx | 4.03B | M | other | giangndm | 1 | 825 |
| sovthpaw/Omni-Senter-3B | quantized | GGUF | — | — | apache-2.0 | sovthpaw | 9 | 297 |
| zhu-thu-22/GuardReasoner-Omni-3B | finetune | — | 4.70B | M | — | zhu-thu-22 | 0 | 661 |
| Memories-ai/UGC-VideoCaptioner | finetune | — | 5.54B | M | mit | Memories-ai | 3 | 494 |
| sj5430/Qwen2.5-Omni-3B-Q4_K_M-GGUF | quantized | GGUF | — | — | other | sj5430 | 1 | 530 |
| zhaoweiguo/Qwen2.5-Omni-3B-Q4_K_M-GGUF | quantized | GGUF | — | — | other | zhaoweiguo | 0 | 518 |
| rfisdhef/Qwen2.5-Omni-3B-GGUF | quantized | GGUF | — | — | other | rfisdhef | 0 | 482 |
| shlv/AudioJev | finetune | — | 4.70B | M | other | shlv | 4 | 274 |
| Jazzcharles/AuroLA-Omni-3B | finetune | — | 4.70B | M | — | Jazzcharles | 3 | 207 |
| mlx-community/Qwen2.5-Omni-3B-Thinker-4bit | finetune | -4bit | — | — | other | mlx-community | 1 | 257 |
| sugiv/octopus-omni-embed | finetune | — | 5.54B | M | apache-2.0 | sugiv | 1 | 251 |
| RUC-NLPIR/OmniAtlas-Qwen2.5-3B | finetune | — | 5.54B | M | apache-2.0 | RUC-NLPIR | 1 | 234 |
| mlx-community/Qwen2.5-Omni-3B-Thinker-8bit | finetune | -8bit | — | — | other | mlx-community | 1 | 206 |
| vidore/colqwen2.5omni-base | finetune | — | 4.39B | M | other | vidore | 0 | 250 |
| cocolati/Qwen2.5-Omni-3B-Q4_K_M-GGUF | quantized | GGUF | — | — | other | cocolati | 0 | 233 |
| namin0202/qwen2-5-omni-3b_video-onecall-ours | adapter | — | — | — | — | namin0202 | 0 | 218 |
| namin0202/qwen2-5-omni-3b_video-onecall-caption-ours | adapter | — | — | — | — | namin0202 | 0 | 208 |
| RKNNAI/RK182X-MLLM-Qwen2.5-Omni-3B | quantized | GGUF | — | — | other | RKNNAI | 0 | 201 |
| epfl-neuroai/topo-omni | finetune | — | 4.96B | M | apache-2.0 | epfl-neuroai | 0 | 170 |
| MYJOKERML/SDiaReward-3B | finetune | — | 4.70B | M | apache-2.0 | MYJOKERML | 0 | 156 |
| eduardo-terres2/Wianmu-3B | finetune | — | — | — | apache-2.0 | eduardo-terres2 | 2 | 41 |
| weihongliang/jevtrainer-omni3b-2026-10-04 | finetune | — | — | — | apache-2.0 | weihongliang | 0 | 139 |
| andrewt28/qwen2.5-omni-3b-keyboard-video-text | adapter | — | — | — | afl-3.0 | andrewt28 | 0 | 123 |
| namin0202/qwen25-omni-3b-star-iter3-ours | adapter | — | — | — | — | namin0202 | 0 | 103 |
| TeenageWorks/Toddler-Jev | adapter | — | — | — | other | TeenageWorks | 1 | 51 |
| liangyuch/ttcc-sft-qwen25omni-3b-v8 | finetune | — | 1.26M | S | other | liangyuch | 0 | 74 |
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.
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).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:37 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
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
