itsZyn/ZynDwarf-1.1

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.35BQuant:BF16Context Size:32kPublished:Sep 12, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

ZynDwarf-1.1 is a compact 354.5 million parameter text-only general assistant model developed by Zyn Models / itsZyn, derived from Liquid AI's LFM2.5-350M family. Optimized for local deployment on CPU hardware, it excels in everyday conversation, Spanish and English interaction, programming, debugging, and agent-oriented workflows with native tool-call formatting. This model is designed for practical reasoning and structured output, supporting efficient integration into local tools and agent runtimes.

Loading preview...

ZynDwarf-1.1: A Compact Agent-Oriented Model

ZynDwarf-1.1, developed by Zyn Models / itsZyn, is a 354.5 million parameter language model built upon Liquid AI's LFM2.5-350M. It is specifically designed for efficient local deployment, targeting CPU-oriented hardware and embedded applications. The model's core differentiator is its focus on agent-oriented workflows, enabling it to decide when a tool is useful, emit structured tool calls, and integrate seamlessly into host-controlled agent loops.

Key Capabilities

  • Bilingual Support: Proficient in everyday conversation and explanations in both Spanish and English.
  • Programming & Debugging: Handles Python, JavaScript, TypeScript, Bash, HTML/CSS, JSON, and SQL tasks, including debugging and code transformation.
  • Agentic Behavior: Supports tool selection, tool-call formatting, tool-result continuation, recovery from tool failures, and multi-turn agent interactions, with a native tool-call template (<|tool_call_start|>[ToolName(arg=value)]<|tool_call_end|>).
  • Structured Output: Designed to produce structured responses for various applications.
  • Local Inference Optimized: Provided in Safetensors, F16 GGUF, and Q4_K_M GGUF formats, with Q4_K_M offering a 68% size reduction and faster CPU throughput for resource-constrained environments.
  • Context Length: Features an architectural maximum position length of 128,000 tokens, with a recommended runtime context of 32,768 tokens.

Ideal Use Cases

  • Local CPU Deployment: Excellent for applications requiring a small footprint and efficient execution on standard CPUs.
  • Autonomous Agents: Suitable for small autonomous or semi-autonomous agents where the host application manages security and tool execution.
  • Developer Tools: Can be embedded into local development tools for coding assistance, debugging, and structured output generation.
  • Multilingual Assistants: Effective for conversational agents needing to operate in Spanish and English.