Search the full model catalog.

Sings new lyrics in the style of a song you upload. YuE's in-context mode uses the first 30 seconds of the reference's vocal stem, backing stem or both as the prompt.

Sings your section-labeled lyrics as a track with vocals and accompaniment, using the open YuE models. Required genre, instrument, mood, gender and timbre tags set the style.

Generates instrumental music from a text description with Meta's MusicGen Large. A token budget sets clip length, about 20 seconds at the default, and guidance sets how closely it follows the prompt.

SkyReels V1 is the first and most advanced open-source human-centric video model.

DeepSeek's mixture-of-experts vision-language model, run here as the Tiny checkpoint. Give it one image and a question and it answers in plain text, from a full caption to a one-line note on what looks odd.

Video NSFW Detection checks a video for explicit content and returns one label, normal or nsfw, using one of two open ViT image classifiers.

Stability AI's 8.1B-parameter MMDiT text-to-image model from the Stable Diffusion 3.5 release, built for 1 megapixel images in styles that range from photography and 3D to painting and line art.

DAMO-NLP-SG's 2B VideoLLaMA 3 model watches a video clip and answers in English text. It skips image patches that stay the same from frame to frame, so static shots use fewer tokens.

The 2B image checkpoint of DAMO-NLP-SG's VideoLLaMA 3 answers questions about a picture in English. Larger images get more vision tokens, which helps with small print in documents and charts.

Turns a few lines of hiring notes into a formatted job posting in English or Turkish, in one of five tones. CVs or an older posting can be attached as extra context.

MEMO animates a single portrait so it talks or sings along to an audio clip, keeping the face consistent over long videos and matching the expression to the emotion in the voice.

Qwen2.5-14B-Instruct is the 14.7B instruction-tuned model from Alibaba's Qwen2.5 series, built for chat, multilingual writing, coding help and JSON output.

Qwen reports 83.1 on MATH and 74.5 on Arena-Hard for this 32.5B-parameter Qwen2.5 chat model. On Wiro a request holds about 17,000 input tokens, with JSON mode and function calling.

Sana 1.6B adapted by NVIDIA for 2K text-to-image output: images around 2048 × 2048 in BF16, from prompts in English, Chinese, emoji or a mix.

NVIDIA's 1.6B-parameter Sana text-to-image model in BF16, built for 1024px images. It accepts English, Chinese and emoji prompts, and the 2K and 4K Sana checkpoints were fine-tuned from it.

Half-precision (FP16) weights for Sana 1.6B at 1024px. NVIDIA's model pairs a linear diffusion transformer with a Gemma 2 text encoder and a 32x compression autoencoder.

The 4K checkpoint of NVIDIA's 1.6B Sana model, fine-tuned from the 1024px BF16 release to generate images up to 4096 × 4096, with prompts in English, Chinese or emoji.

The largest Qwen-based R1 distill, built by DeepSeek on Qwen2.5-32B. DeepSeek's results put it ahead of OpenAI o1-mini on AIME 2024, MATH-500, GPQA Diamond and LiveCodeBench.

The mid-sized R1 distill: Qwen2.5-14B fine-tuned on DeepSeek-R1 samples. DeepSeek reports results close to or above OpenAI o1-mini on most of the benchmarks it lists.

DeepSeek's R1 distill on a Llama 3.1 8B base. It writes out a chain of thought before answering and is the only Llama option among the smaller R1 distills.