OmniJev/OneJev1.5-27B
OmniJev/OneJev1.5-27B is a 27 billion parameter multimodal System One decision model developed by OmniJev, fine-tuned from Qwen/Qwen3.8-27B. It processes screenshots, photos, videos, and text inputs to return calibrated probabilities for decision options in a single pass. This model excels in decision-making tasks, achieving the highest Arts & Human Taste score and second-highest Language Understanding score on the Decision Index 0.3 benchmark, making it suitable for complex multimodal decision support.
Loading preview...
Overview
OneJev1.5-27B is a 27 billion parameter multimodal System One decision model developed by OmniJev. It is a full fine-tune of Qwen/Qwen3.8-27B, trained on approximately 320,000 questions derived from real agent runs, videos, images, and public decision datasets. The model is designed to take various inputs, including screenshots, photos, videos, or plain text, along with specific questions, and output calibrated probabilities for each available option in a single forward pass.
Key Capabilities
- Multimodal Decision Making: Processes diverse inputs like images, videos, and text to inform decisions.
- Calibrated Probabilities: Provides a probability score for each decision option, aiding in nuanced decision support.
- High Performance on Decision Index: Achieves a score of 63.64 on the public part of Decision Index 0.3, ranking fifth among 148 models.
- Specialized Strengths: Holds the highest score in Arts & Human Taste and the second-highest in Language Understanding on the Decision Index.
Good for
- Automated agent decision support systems requiring multimodal input.
- Applications needing probabilistic outputs for complex choices.
- Tasks involving visual and textual understanding for decision-making, such as analyzing screenshots or video frames.
- Use cases where high accuracy in human taste and language comprehension is critical for decision outcomes.