Up to date
Qwen Apache 2.0 out 23 months ago
32.8B parameters, text in, text out, 32K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #199 in LLMs · score 88.9 method v1 Sources 3 Downloads 30d 9.1k · #514 of 619 tracked
Picked by rule from 22 tracked pieces: the lab's own post, the most shared, a field report, the latest.
QwQ-32B-Preview sometimes switches between English and Chinese, which complicates human understanding of its outputs.
QwQ-32B is a 32 billion parameter model available under the Apache 2.0 license that matches DeepSeek-R1's performance despite having only 32 billion parameters versus 671 billion.
Ubicloud launched inference APIs for the open source QwQ-32B-Preview model, which offers reasoning capabilities comparable to OpenAI's o1 at significantly lower cost.
Alibaba deprecated QwQ 32B and suggests considering the newer Qwen3 32B instead.
Unsloth provides a guide for running QwQ-32B effectively, addressing issues like infinite generations and repetition.
The 4 picks above, plus these 18. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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 | Featherless | — | — | — | — | Oct 25 |
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 (105 quantized builds); a big base for builders (111 fine-tunes, adapters and merges).
Top Spaces running it: llamameta/google-gemini-web-search ♥ 24 · llamameta/Fix-qwen-QwQ-32B-Preview-improvement ♥ 11 · siddqamar/Exam-Preparation-Assistant ♥ 8 · suayptalha/QwQ-32B-Preview-Vision ♥ 7
Weights only, an estimate: leave room for the KV cache. Method
About 20 GB at INT4, 39 GB at FP8, 79 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 (2024-11-27), amber triangles are 9 dated articles (each links to it), grey ticks are repo commits.
216 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.
It supports a context length of 32,768 tokens.[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 | text-generation |
|---|---|
| Library | transformers |
| Architectures | Qwen2ForCausalLM |
| Model type | qwen2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 32,763,876,352 |
| Size band | L |
| MoE | no |
| Context length | 32,768 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2407.10671 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2ForCausalLM · 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 QwQ-32B-Preview: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 79 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 39 GB · Single high-end GPU (<80 GB) |
| INT4 (0.5 bytes/param) | 20 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: Qwen/Qwen2.5-32B-Instruct
quantized: 105finetune: 38merge: 48adapter: 25
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| KirillR/QwQ-32B-Preview-AWQ | quantized | AWQ | 32.8B | L | apache-2.0 | KirillR | 24 | 543.7k |
| bartowski/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | bartowski | 102 | 259.7k |
| karakuri-ai/karakuri-lm-32b-thinking-2501-exp | finetune | — | 32.8B | L | apache-2.0 | karakuri-ai | 9 | 14.3k |
| lmstudio-community/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | lmstudio-community | 16 | 14k |
| huihui-ai/QwQ-32B-Preview-abliterated | finetune | — | 32.8B | L | apache-2.0 | huihui-ai | 103 | 8.2k |
| unsloth/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | unsloth | 12 | 12.5k |
| unsloth/QwQ-32B-Preview-bnb-4bit | quantized | -bnb | 33.7B | L | apache-2.0 | unsloth | 4 | 12k |
| mradermacher/QwQ-32B-Preview-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 1 | 10.4k |
| mmnga/QwQ-32B-Preview-gguf | quantized | gguf | — | — | apache-2.0 | mmnga | 3 | 9.7k |
| nanowell/QwQ-32B-Preview-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | nanowell | 7 | 6.1k |
| featherless-ai-quants/Qwen-QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | — | featherless-ai-quants | 3 | 5.9k |
| unsloth/QwQ-32B-Preview-unsloth-bnb-4bit | quantized | -bnb | 33.7B | L | apache-2.0 | unsloth | 19 | 4k |
| ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v1 | quantized | gptq | 32.8B | L | apache-2.0 | ModelCloud | 51 | 1.9k |
| DevQuasar-5/Qwen.QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | — | DevQuasar-5 | 4 | 3.5k |
| second-state/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | second-state | 0 | 3.5k |
| bartowski/QwQ-32B-Preview-exl2 | quantized | exl2 | — | — | apache-2.0 | bartowski | 13 | 2.8k |
| ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v2 | quantized | gptq | 32.8B | L | apache-2.0 | ModelCloud | 16 | 2.3k |
| gaianet/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | gaianet | 0 | 2k |
| nitky/RoguePlanet-DeepSeek-R1-Qwen-32B | merge | — | 32.8B | L | apache-2.0 | nitky | 8 | 1.6k |
| mradermacher/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 0 | 1.9k |
| sbeltz/QwQ-32B-Preview-Q3_K_S-GGUF | quantized | GGUF | — | — | apache-2.0 | sbeltz | 3 | 1.7k |
| ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v3 | quantized | gptq | 32.8B | L | apache-2.0 | ModelCloud | 14 | 1.1k |
| tensorblock/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | tensorblock | 0 | 1.8k |
| mlx-community/QwQ-32B-Preview-4bit | quantized | -4bit | 32.8B | L | apache-2.0 | mlx-community | 3 | 1.5k |
| prithivMLmods/Qwen2.5-32B-DeepSeek-R1-Instruct | merge | — | 32.8B | L | — | prithivMLmods | 4 | 1.5k |
| OpenBuddy/openbuddy-qwq-32b-v24.1-200k | finetune | — | 32.8B | L | apache-2.0 | OpenBuddy | 3 | 1.5k |
| mlx-community/QwQ-32B-Preview-8bit | quantized | -8bit | 32.8B | L | apache-2.0 | mlx-community | 6 | 1.3k |
| unsloth/QwQ-32B-Preview | finetune | — | 32.8B | L | apache-2.0 | unsloth | 7 | 1.1k |
| Sakalti/oxyge1-33B | merge | — | 32.8B | L | apache-2.0 | Sakalti | 2 | 1.3k |
| prithivMLmods/Omni-Reasoner4-Merged | merge | — | 32.8B | L | — | prithivMLmods | 2 | 1.2k |
| Ba2han/QwQenSeek-coder | merge | — | 32.8B | L | — | Ba2han | 5 | 874 |
| mlx-community/QwQ-32B-Preview-3bit | quantized | — | 32.8B | L | apache-2.0 | mlx-community | 5 | 830 |
| wanlige/QWQ-stock | merge | — | 32.8B | L | — | wanlige | 8 | 680 |
| mlx-community/Qwen_QwQ-32B-Preview_MLX-8bit | quantized | -8bit | 32.8B | L | apache-2.0 | mlx-community | 4 | 871 |
| mlx-community/QwQ-32B-Preview-bf16 | finetune | -bf16 | 32.8B | L | apache-2.0 | mlx-community | 5 | 790 |
| mlx-community/QwQ-32B-Preview-6bit | quantized | — | 32.8B | L | apache-2.0 | mlx-community | 4 | 752 |
| divpasta123/QwQ-32B-Preview-GGUF | quantized | GGUF | — | — | apache-2.0 | divpasta123 | 0 | 891 |
| 6cf/QwQ-32B-Preview-IdeaWhiz-v1 | finetune | — | 32.8B | L | apache-2.0 | 6cf | 10 | 281 |
| TeetouchQQ/model_mergev2 | merge | — | 32.8B | L | — | TeetouchQQ | 1 | 730 |
| ehristoforu/0109-test-32b-it | merge | — | 32.8B | L | — | ehristoforu | 15 | 24 |
| Date | Commit | SHA |
|---|---|---|
| 2025-01-12 | Update README.md | 91906fe |
| 2024-11-29 | Update tokenizer_config.json | 1032e81 |
| 2024-11-29 | Update README.md | edf11ce |
| 2024-11-27 | Update README.md | 7e8c9b2 |
| 2024-11-27 | Update README.md | 89beac9 |
| 2024-11-27 | Update README.md | e4c797f |
| 2024-11-27 | Create LICENSE | 8d9c448 |
| 2024-11-27 | Create README.md | fbc0ca8 |
| 2024-11-27 | Upload model | d233ae7 |
| 2024-11-27 | initial commit | db321d2 |
Hosted endpoints observed for QwQ-32B-Preview, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Featherless hf | Qwen/QwQ-32B-Preview | — | — | — | — | — | — | — | live | 2026-10-25 |
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 21:36 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
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
