OmniJev/OneJev1.5-9B
OmniJev/OneJev1.5-9B is a 9-billion parameter multimodal System One decision model developed by OmniJev, fine-tuned from Qwen/Qwen3.5-9B. It processes screenshots, photos, videos, and text inputs to return calibrated probabilities for decision options in a single forward pass. With a 32768-token context length, it excels at decision-making tasks, achieving the highest score (59.03) among models up to 10B parameters on the public part of Decision Index 0.3, and outperforming many larger models.
Loading preview...
Overview
OmniJev/OneJev1.5-9B is a 9-billion parameter multimodal System One decision model, developed by OmniJev. It is a full fine-tune of Qwen/Qwen3.5-9B, trained on approximately 320,000 questions derived from real agent runs, videos, images, and public decision datasets. This model is designed to take various inputs, including screenshots, photos, videos, and plain text, along with specific questions, and output calibrated probabilities for each decision option in a single pass.
Key Capabilities
- Multimodal Decision Making: Processes diverse inputs like images (screenshots, photos), videos, and text for comprehensive decision analysis.
- Calibrated Probabilities: Provides precise, calibrated probabilities for decision options, aiding in automated or assisted decision-making.
- High Performance: Achieved a score of 59.03 on the public part of Decision Index 0.3, making it the top-performing model among those up to 10 billion parameters and surpassing many models in the 25B to 36B parameter range.
Good For
- Automated Agent Systems: Ideal for scenarios requiring agents to make decisions based on visual and textual information.
- Complex Decision Tasks: Suitable for applications needing to evaluate multiple options with probabilistic outcomes from varied data sources.
- Benchmarking Decision Models: Serves as a strong baseline or competitor for decision-making benchmarks, particularly in multimodal contexts.