startlux-models/StartLux-Decision-v1.1-4B-Instant

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 10, 2026License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

StartLux-Decision-v1.1-4B-Instant by StartLux Labs is a 4.5 billion parameter decision model designed to provide probabilistic answers to typed questions (choice, yes/no, score) based on a given state, which can include text or up to 8 images. Unlike generative LLMs, it returns probabilities for options from a single forward pass without generating text, making it efficient for structured decision-making. It supports a 32768 token context length and is optimized for direct decision inference, compatible with the System One /v1/systemone API.

Loading preview...

StartLux-Decision-v1.1-4B-Instant Overview

StartLux-Decision-v1.1-4B-Instant is a 4.5 billion parameter model developed by StartLux Labs, specifically engineered for structured decision-making. It processes a given state (text or JSON, with up to 8 images) and responds to typed questions by providing probabilities for predefined options, rather than generating free-form text. This model supports three question types: choice (pick one of several options), noul (yes or no), and score (a level on a scale of 2 to 10).

Key Capabilities

  • Direct Decision Inference: Returns probabilities for all options in a single forward pass, optimizing for speed and efficiency in decision tasks.
  • Multimodal Input: Accepts both text and up to 8 images as part of the input state for comprehensive context.
  • Structured Output: Provides clear, probabilistic answers for choice, noul, and score question types.
  • API Compatibility: Designed to be compatible with the System One /v1/systemone API request and response formats.
  • Bidirectional Prompt Attention: Utilizes a unique attention mechanism where prompt tokens attend to later prompt tokens in full-attention layers, allowing the model to read the state with the question in view.

Good For

  • Automated Triage and Routing: Ideal for systems requiring rapid, probabilistic decisions, such as customer support ticket routing or content moderation.
  • Structured Data Analysis: When you need to extract specific, quantifiable decisions from complex inputs without generating conversational text.
  • High-Throughput Decision Systems: Its non-generative nature and single-pass inference make it suitable for applications demanding low latency and high concurrency.
  • Research in Decision AI: The model's architecture and performance metrics, detailed in its technical report, offer a valuable resource for studying decision-making in AI.

Finetunes
4 models