CMSManhattan/JiRackUltra_14b

TEXT GENERATIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 3, 2026License:mitArchitecture:Transformer0.0K Open Weights Featherless Exclusive Cold

CMSManhattan/JiRackUltra_14b is a 14.8 billion parameter language model built on a DeepSeek R1-14B architecture, refactored with native ternary (BitNet-style) support and an updated tokenizer. Optimized for CPU inference, it features specialized tags for Routing, Media, Vision, Sound, Tool call, and Robotics. This model is designed for efficient cloud deployments and serves as an expert model in RAG systems, offering ready-to-run GGUF quantizations for various performance needs.

Loading preview...

JiRack Ultra 14B: CPU-Optimized Ternary Model with Advanced Tooling

CMSManhattan/JiRackUltra_14b is a 14.8 billion parameter model built on a DeepSeek R1-14B architecture, uniquely refactored with native ternary (BitNet-style) support for high-quality CPU inference. It features an updated tokenizer that includes specialized tags for Routing, Media, Vision, Sound, Tool call, and Robotics, making it highly adaptable for complex, domain-specific applications.

Key Capabilities

  • CPU-Optimized Performance: Designed for efficient inference on CPUs, supporting GGUF quantizations (Q2_K, Q3_K_M, Q4_K_M) for various memory and speed requirements.
  • Ternary Architecture: Incorporates BitNet features for enhanced compression and efficiency, with options for custom Quantization-Aware Training (QAT) for specific datasets.
  • Advanced Tool Calling: The specialized JiRack Precision Tokenizer enables high-quality tool calling, making it suitable for agentic and instruct models.
  • Cloud-Ready & Cost-Effective: Optimized for cloud infrastructure, reducing operational costs, and can function as an expert model in RAG deployments.
  • Integrated Development Environment (IDE) Support: Offers plugins for Eclipse, JetBrains, and VS Code, facilitating agent coding with JiRack models.

Good For

  • Cost-Sensitive Deployments: Ideal for users seeking powerful LLM capabilities without extensive GPU requirements, leveraging CPU efficiency.
  • Specialized AI Agents: Excellent for developing agents requiring advanced routing, multimedia processing, robotics control, or precise tool invocation.
  • RAG Systems: Can be effectively deployed as an expert model within Retrieval-Augmented Generation architectures.
  • Edge Computing: Suitable for scenarios where computational resources are limited, such as on-device or edge deployments, due to its CPU optimization and compact quantizations.

Finetunes
1 models