Decision-1
Worthify’s Decision-1 model family: the original 12B preview, smaller 0.6B/E2B/E4B releases, NF4 quantization, and BANKING77 LoRA.
Text Generation • 12B • Updated • 945 • 4Note Original 12B full-weight, single-seed research preview. Scores 2–16 supplied options; conditional scores are uncalibrated.
Worthify/Decision-1-0.6B
Text Generation • 0.6B • Updated • 66 • 1Note Smallest release, based on Qwen3-0.6B. Includes merged Transformers weights, F16 GGUF, and a local Electron/Node scorer.
Worthify/Decision-1-E2B
Text Generation • 5B • Updated • 69 • 1Note Gemma 4 E2B-based release with Transformers weights, F16 GGUF, and a local Electron/Node scorer. Resident text parameters: 4.63B.
Worthify/Decision-1-E4B
Text Generation • 7B • Updated • 79Note Gemma 4 E4B-based release with Transformers weights, F16 GGUF, and a local Electron/Node scorer. Resident text parameters: 7.46B.
Worthify/Decision-1-NF4
Text Generation • 12B • Updated • 23Note Pre-quantized 4-bit NF4 version of Decision-1. Adds no training; quantization can change selected options. Follow its numeric policy.
Worthify/decision-1-banking77-lora
Updated • 40Note BANKING77 task LoRA for the pinned Decision-1 base. Evaluated with 16-option menus; see the card for the matched Gemma control.