Up to date
Qwen Qwen License (qwen) out 22 months ago
73.4B parameters, images and text in, text out, 125K tokens of context.
Verified 2026-10-25 Rank #436 in LLMs · score 78.4 method v1 Sources 3 Downloads 30d 1.4k · #586 of 619 tracked
Picked by rule from 2 tracked pieces: the lab's own post, the most shared, a field report, the latest.
QVQ-72B-Preview is an open-weight multimodal reasoning model built upon Qwen2-VL-72B that achieves a score of 70.3 on the MMMU benchmark.
QvQ-72B-Preview is a vision-focused follow-up to QwQ that uses a long chain-of-thought process to solve visual problems.
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: 24 quantized builds and 36 Spaces.
Top Spaces running it: evbro/Qwen-QVQ-72B-Preview99 ♥ 2 · hrsprojects/R ♥ 1 · OP7/SG_TestSpace ♥ 1
Weights only, an estimate: leave room for the KV cache. Method
About 44 GB at INT4, 88 GB at FP8, 176 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 (2024-12-24), amber triangles are 2 dated articles (each links to it), grey ticks are repo commits.
30 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 MMMU · Understand images and video | 70.3 | — | model author | 2026-10-07 |
| author-reported MathVision · Understand images and video | 35.9 | — | model author | 2026-10-07 |
| author-reported MathVista · Understand images and video | 71.4 | — | model author | 2026-10-07 |
| author-reported OlympiadBench | 20.4 | — | model author | 2026-10-07 |
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 | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | Qwen2VLForConditionalGeneration |
| Model type | qwen2_vl |
| Dtype | bfloat16 |
| Params by dtype | BF16: 73,405,560,320 |
| Size band | XL |
| MoE | no |
| Context length | 128,000 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2409.12191 |
| License | Qwen License (qwen) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2VLForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2VLForConditionalGeneration · 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 QVQ-72B-Preview: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 176 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 88 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.
Base model: Qwen/Qwen2-VL-72B
quantized: 24finetune: 3adapter: 3
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| bartowski/QVQ-72B-Preview-GGUF | quantized | GGUF | — | — | other | bartowski | 56 | 235.5k |
| unsloth/QVQ-72B-Preview-bnb-4bit | quantized | -bnb | 75.6B | XL | other | unsloth | 5 | 18.3k |
| kosbu/QVQ-72B-Preview-AWQ | quantized | AWQ | 73.7B | XL | other | kosbu | 8 | 9k |
| mmnga/QVQ-72B-Preview-gguf | quantized | gguf | — | — | other | mmnga | 0 | 8.2k |
| mradermacher/QVQ-72B-Preview-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 7.4k |
| lmstudio-community/QVQ-72B-Preview-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 2 | 5.3k |
| tensorblock/QVQ-72B-Preview-GGUF | quantized | GGUF | — | — | other | tensorblock | 0 | 3.2k |
| mradermacher/QVQ-72B-Preview-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 2.7k |
| DevQuasar-6/Qwen.QVQ-72B-Preview-GGUF | quantized | GGUF | — | — | — | DevQuasar-6 | 3 | 2.1k |
| matatonic/QVQ-72B-Preview-abliterated-6.5bpw-h8-exl2 | quantized | 5bpw | — | — | other | matatonic | 1 | 211 |
| OPEA/QVQ-72B-Preview-int4-sym-inc | quantized | int4 | 12.6B | L | — | OPEA | 0 | 249 |
| shad777725/risk_model | finetune | — | — | — | apache-2.0 | shad777725 | 0 | 142 |
| Dracones/QVQ-72B-Preview_exl2_2.75bpw | quantized | exl2 | — | — | other | Dracones | 0 | 133 |
| bullerwins/QVQ-72B-Preview-exl2_4.5bpw | quantized | exl2 | — | — | other | bullerwins | 0 | 133 |
| Dracones/QVQ-72B-Preview_exl2_8.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 129 |
| Dracones/QVQ-72B-Preview_exl2_2.25bpw | quantized | exl2 | — | — | other | Dracones | 0 | 69 |
| Dracones/QVQ-72B-Preview_exl2_6.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 61 |
| Dracones/QVQ-72B-Preview_exl2_3.5bpw | quantized | exl2 | — | — | other | Dracones | 0 | 61 |
| bullerwins/QVQ-72B-Preview-exl2_5.0bpw | quantized | exl2 | — | — | other | bullerwins | 0 | 59 |
| Dracones/QVQ-72B-Preview_exl2_5.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 58 |
| Dracones/QVQ-72B-Preview_exl2_3.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 56 |
| Dracones/QVQ-72B-Preview_exl2_4.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 55 |
| Dracones/QVQ-72B-Preview_exl2_4.5bpw | quantized | exl2 | — | — | other | Dracones | 0 | 54 |
| Dracones/QVQ-72B-Preview_exl2_7.0bpw | quantized | exl2 | — | — | other | Dracones | 0 | 53 |
| Dracones/QVQ-72B-Preview_exl2_2.5bpw | quantized | exl2 | — | — | other | Dracones | 0 | 42 |
| sangeetha1993/Coder-Qwen | finetune | — | — | — | mit | sangeetha1993 | 0 | 0 |
| DDBMST3W/OLE_ENG | adapter | — | — | — | mit | DDBMST3W | 0 | 0 |
| whitedell/YourAi | finetune | — | — | — | apache-2.0 | whitedell | 0 | 0 |
| rahmatov33/lotheus | adapter | — | — | — | — | rahmatov33 | 0 | 0 |
| Jaythakur/clothify | adapter | — | — | — | apache-2.0 | Jaythakur | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2025-01-12 | Update README.md | 24425f6 |
| 2024-12-25 | Update LICENSE | 53b19b9 |
| 2024-12-24 | Update preprocessor_config.json | a45049b |
| 2024-12-24 | Update README.md | decb20e |
| 2024-12-24 | Create LICENSE | 5190b32 |
| 2024-12-24 | Update README.md | 352ec82 |
| 2024-12-24 | Update README.md | 5724f82 |
| 2024-12-24 | Create README.md | 55fe770 |
| 2024-12-24 | Update chat_template.json | fc9b0c2 |
| 2024-12-24 | Upload folder using huggingface_hub | e5d3a0d |
| 2024-12-24 | initial commit | 980ce2b |
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 21:40 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:12 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.
