this-that-model-1.2
One forward pass, no decoding: calibrated typed decisions
None defined yet.
One forward pass, no decoding: calibrated typed decisions
Multilingual zero-shot text classification, any label set
One checkpoint, any label set, instant text decisions
Box an object and let Qwen-Image-2.1 erase it.
Canny/depth/pose/β¦ control + inpainting for Qwen-Image 2.1
Segment 60 anatomical structures from any clinical X-ray
Single-pass typed decisions with calibrated probabilities
Swedish ASR with Klang Pianissimo, word timestamps
Typed decisions with calibrated probabilities, one pass
Native streaming Chinese/English ASR, 80ms clock
Physics-guided ULF (0.064T) to 1.5T MRI enhancement
Few-step (10-32) distilled protein backbone generation
Fine-grained visual perception VLM (images & video)
Decompose a chest X-ray into bone and soft-tissue images
Next GUI action or MCP tool call from a screenshot
One forward pass, no decoding: calibrated typed decisions
Structured document parsing with a 0.8B VLM
Read long docs as compressed page images, expand on demand
SNES-era pixel-art video with sound, from text
Transcribe audio with moondream/parakeet-ultra
State in, a probability distribution out. Nothing decoded.
Classify & explain memes across 46 tasks in 9 languages
Configurable multi-stage crop disease and pest diagnosis
Epic orchestral trailer songs with the YuE2 TSFH LoRA