interpolators/FableOpus-9B-Delta

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 21, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

FableOpus-9B-Delta is a 9 billion parameter language model developed by interpolators, built on the Qwen3.5-9B architecture. This model is a delta merge designed to combine the agentic and tool-use capabilities of the Fable 5 distillation with the strong reasoning performance of Claude Opus-distilled checkpoints. It is optimized for general chat, code assistance, tool-use style prompting, and reasoning-heavy tasks, supporting a 32768 token context length.

Loading preview...

FableOpus-9B-Delta: Merging Agentic Tool-Use with Opus Reasoning

FableOpus-9B-Delta is a 9 billion parameter model built upon the Qwen3.5-9B base, developed by interpolators. This model represents a strategic delta merge, combining distinct strengths from multiple source signals to create a versatile and powerful LLM. It specifically integrates the agentic and tool-use behaviors derived from Fable 5 distillation with the high-performance reasoning capabilities found in Claude Opus-distilled checkpoints.

Key Capabilities

  • Hybrid Intelligence: Blends the practical, agentic prompting style of Fable 5 with the advanced reasoning of Opus-distilled models.
  • Qwen3.5-9B Family: Benefits from the robust architecture and performance characteristics of the Qwen3.5-9B base model.
  • Bfloat16 Precision: Utilizes bfloat16 for efficient computation and memory usage.
  • Tool-Use Integration: Designed to maintain and enhance tool-use style prompting, making it suitable for agentic workflows.

Good For

  • General Chat Applications: Capable of engaging in diverse conversational tasks.
  • Code Assistance: Provides support for coding-related queries and generation.
  • Reasoning-Heavy Experiments: Excels in tasks requiring complex logical thought and problem-solving.
  • Agentic Workflows: Ideal for scenarios where models need to interact with tools or follow multi-step instructions.