Laya answers typed decisions (choose one option, give a score, answer yes/no) with calibrated probabilities. laya-omni adds optional image and audio inputs. Without an image or audio clip, the output is bit-identical to Laya. Qwen3-ASR-0.6B audio encoder extracted from the full ASR model (audioencoder/). base/16-256 (~93M), Qwen3-ASR audio encoder (186M). Several images: image=[a, b]. detail=True uses 256 image tokens instead of 64 (better for text, documents and spatial questions). One question with an image or a 10 s clip takes 30-120 ms on an RTX 4090, encoders included. Standard benchmarks, same questions, one RTX 4090, one question at a time from the image or audio file to the answer…
Independent publisher
Zheqiushui
zheqiushui
Models in Library1
Datasets in Library0
Models on Hugging Face2
Followers1
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Any to any.
Reported licenses: other. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.