Up to date
deepseek-ai MIT (deepseek) out 21 months ago
Several kinds of input and output.
Verified 2026-10-25 Rank #196 in LLMs · score 89.3 method v1 Sources 3 Downloads 30d 10.2k · #504 of 619 tracked
Changed this week: 1 new quantized build on the Hub (4 in all).2026-10-24
Picked by rule from 15 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Janus is a single Go binary that runs .gguf models on your machine and exposes an OpenAI-compatible API.
The 2 picks above, plus these 13. 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: 4 quantized builds and 74 Spaces.
Top Spaces running it: deepseek-ai/Janus-Pro-7B ♥ 2,035 · VIDraft/Deepseek-Multimodal ♥ 118 · AP123/Janus-Pro-7b ♥ 109 · LLMhacker/DeepseekJanusPro-Image ♥ 63
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2025-01-26), amber triangles are 6 dated articles (each links to it), grey ticks are repo commits.
49 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.
Janus-Pro-7B is an any-to-any multimodal model that unifies multimodal understanding and generation.[1][2] The architecture utilizes a single transformer to process decoupled visual encoding pathways.[3]
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 |
| Model type | multi_modality |
| Dtype | bfloat16 |
| MoE | no |
| Papers | arXiv:2501.17811 |
| License | MIT (deepseek) (mit) |
| Hub snapshot | 2026-10-25 |
SGLang: supported · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
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.
finetune: 37quantized: 4adapter: 8
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Franklin0/ReasonGen-R1-SFT | finetune | — | 7.42B | M | apache-2.0 | Franklin0 | 1 | 14.7k |
| FreedomIntelligence/Janus-4o-7B | finetune | — | 7.42B | M | mit | FreedomIntelligence | 52 | 5.1k |
| Franklin0/ReasonGen-R1 | finetune | — | 7.42B | M | apache-2.0 | Franklin0 | 7 | 5k |
| wnma3mz/Janus-Pro-7B-4bit | quantized | -4bit | 7.42B | M | mit | wnma3mz | 8 | 3.4k |
| CaraJ/T2I-R1 | finetune | — | 7.42B | M | apache-2.0 | CaraJ | 5 | 1.7k |
| fibonacciai/Persian-llm-fibonacci-1-7b-chat.P1_0 | adapter | — | — | — | mit | fibonacciai | 29 | 0 |
| wnma3mz/Janus-Pro-7B-LM-4bit | quantized | -4bit | 6.91B | M | mit | wnma3mz | 2 | 1.1k |
| wnma3mz/Janus-Pro-7B | finetune | — | 7.42B | M | mit | wnma3mz | 0 | 926 |
| wnma3mz/Janus-Pro-7B-LM | finetune | — | 6.91B | M | mit | wnma3mz | 3 | 472 |
| Zhare-AI/janus-pro-7b-webgpu | quantized | — | — | — | mit | Zhare-AI | 2 | 396 |
| ZYT0316/Janus-CXR-7B | finetune | — | 7.42B | M | apache-2.0 | ZYT0316 | 4 | 196 |
| trancoso-cc/gibberish-detector | finetune | — | 67.0M | S | mit | trancoso-cc | 0 | 243 |
| MISHANM/deepseek-ai_janus-Pro-7B-fp16 | finetune | -fp16 | 7.42B | M | — | MISHANM | 2 | 3 |
| MatiVexx/Roxi | finetune | — | — | — | artistic-2.0 | MatiVexx | 2 | 0 |
| zeromodels/janus_pro_7b | finetune | — | — | — | mit | zeromodels | 0 | 88 |
| Mubeiya/jarvis-janus-prompt-lora | adapter | — | — | — | — | Mubeiya | 0 | 86 |
| zghhui/Janus-Pro-7B-GCPO-HPS | finetune | — | 7.42B | M | — | zghhui | 0 | 71 |
| KonstantinosKK/Janus-Pro-7B-GCPO | finetune | — | 7.42B | M | other | KonstantinosKK | 0 | 65 |
| masterthor90/DarkDevil | finetune | — | — | — | wtfpl | masterthor90 | 1 | 0 |
| Helen0811/test_repo | finetune | — | — | — | apache-2.0 | Helen0811 | 1 | 0 |
| Kevein/MakeIdea | finetune | — | — | — | cc-by-sa-4.0 | Kevein | 1 | 0 |
| asats/thumbnail-vlm-janus-pro | finetune | — | — | — | mit | asats | 1 | 0 |
| Billyshears/Janus_pro_finetune | finetune | — | 7.42B | M | — | Billyshears | 0 | 50 |
| exdysa/Janus-Pro-7B-SAFETENSORS | finetune | — | 7.42B | M | mit | exdysa | 0 | 44 |
| ORI-Muchim/Janus-Pro-7B-Korean-LoRA | adapter | — | — | — | mit | ORI-Muchim | 0 | 38 |
| danyalmalik/Janus-Pro-7B-ONNX-WebGPU | quantized | — | — | — | other | danyalmalik | 0 | 16 |
| ErenMil/sofa | finetune | — | — | — | openrail | ErenMil | 0 | 0 |
| tulsikopoi/NepAI | finetune | — | — | — | openrail | tulsikopoi | 0 | 0 |
| Zeariaya/DeepSeek | adapter | — | — | — | mit | Zeariaya | 0 | 0 |
| GgDamn/Testees | finetune | — | — | — | apache-2.0 | GgDamn | 0 | 0 |
| elmasryamr/King | finetune | — | — | — | afl-3.0 | elmasryamr | 0 | 0 |
| jasongarrison/JesusBro | adapter | — | — | — | apache-2.0 | jasongarrison | 0 | 0 |
| Vito13/Death | finetune | — | — | — | openrail | Vito13 | 0 | 0 |
| Ferhat591/Kumsal | finetune | — | — | — | apache-2.0 | Ferhat591 | 0 | 0 |
| Smartph90/Activities | finetune | — | — | — | apache-2.0 | Smartph90 | 0 | 0 |
| JacobLasher/pro | adapter | — | — | — | mit | JacobLasher | 0 | 0 |
| Vezel/Data | finetune | — | — | — | mit | Vezel | 0 | 0 |
| Mandeepmk/Try | finetune | — | — | — | apache-2.0 | Mandeepmk | 0 | 0 |
| vilho/gtyfr65 | finetune | — | — | — | apache-2.0 | vilho | 0 | 0 |
| komalkaur/aysha | finetune | — | — | — | apache-2.0 | komalkaur | 0 | 0 |
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.
