DownloadRune 26B-A4B v3 is an open-weight multimodal decision model developed by Invergent and distributed through Surogate, designed for fast classification, scoring, and probability estimation. Based on Gemma 4 26B-A4B, it contains 26.5B parameters with eight of 128 experts activated per token. The model accepts text, structured data, and images within a 262K-token context window and supports three decision formats: Choice for categorical selection, Noul for binary judgments, and Score for ordinal evaluation. It returns probability distributions in a single forward pass, with optional reasoning for uncertain questions. Rune v3 achieved a 57.44 Decision Index score and 83.4% accuracy on a custom 1,196-item image evaluation. Temperature calibration improves the reliability of its confidence estimates. Released under Apache 2.0, it supports inference through Surogate and Transformers, making it suitable for automated classification, visual assessment, and decision workflows.
Features
- 26.5B-parameter Mixture-of-Experts architecture
- 128 experts with eight activated per token
- 262K-token context window for text and structured data
- Native text and image-based decision classification
- Choice, Noul, and Score structured decision formats
- Calibrated probability outputs with adjustable temperature
- Optional reasoning for low-confidence decisions
- Apache 2.0 licensing with Surogate and Transformers support
Project Samples
