frontier-infra/jebadiah-27b

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 26, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Jebadiah 27B by Frontier Infra is a 27 billion parameter System One style decision model based on Qwen3.8-27B, designed to answer typed questions with probabilities over option labels rather than generating text. It supports choice, noul (yes/no), and score question types, processing one forward pass per question. This model excels at structured decision-making tasks, achieving a Decision Index score of 54.67 (No. 5 of 70 open models) and 0.866 accuracy on JevBench v1.4.2 public items, making it suitable for applications requiring probabilistic, type-safe judgments.

Loading preview...

Jebadiah 27B: A System One Decision Model

Jebadiah 27B, developed by Frontier Infra, is a 27 billion parameter model built on the Qwen/Qwen3.8-27B base. Unlike traditional LLMs that generate free-form text, Jebadiah is a "System One" style decision model. It provides probabilistic answers over predefined option labels for typed questions, making it highly suitable for structured decision-making and classification tasks. The model processes one forward pass per question, returning a probability distribution rather than generative text.

Key Capabilities

  • Typed Question Answering: Supports three distinct question types:
    • Choice: Selects one option from N possibilities.
    • Noul: Provides a probability for a yes/no statement (P(yes)).
    • Score: Places a state on an ordered rubric.
  • Probabilistic Outputs: Delivers confidence scores and probabilities for each option, enabling nuanced decision-making.
  • TypeSafe-Compatible API: Features a /v1/systemone endpoint, compatible with existing Jev clients.
  • Strong Performance: Achieves a Decision Index 0.2.1 score of 54.67 (ranking No. 5 among 70 open models) and an accuracy of 0.866 on JevBench v1.4.2's 231 public items.
  • Efficient Inference: Designed for single-forward-pass decision-making, making it efficient for specific use cases.

Good For

  • Automated Decision Systems: Ideal for applications requiring structured, probabilistic judgments, such as routing customer support tickets, classifying inputs, or scoring items based on criteria.
  • Structured Classification: Excels in scenarios where outputs need to be constrained to a predefined set of labels or a numerical scale.
  • Integration with Existing Jev Workflows: Seamlessly integrates with systems using the Jev client due to its compatible API.
  • Resource-Constrained Environments: Available in various formats including GGUF and MLX, allowing deployment on diverse hardware, including Apple Silicon.

Finetunes
3 models