flywheel-ai/agency-ops

TEXT GENERATIONConcurrent Unit Cost:3Model Size:35.1BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 20, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

The flywheel-ai/agency-ops model is a fine-tuned Qwen3.6-35B-A3B variant developed by Flywheel by OpSpot, specifically optimized for the agency-ops domain. This model functions as an AI-employee or advisor for AI-automation agencies, specializing in tasks like lead generation, client onboarding, workflow building, and agency management. It is available in safetensors (~65GB) and GGUF (~20GB) formats, providing a specialized vertical AI solution.

Loading preview...

Overview

The flywheel-ai/agency-ops model is an open-source vertical AI-employee developed by Flywheel by OpSpot. It is a LoRA fine-tune of the Qwen/Qwen3.6-35B-A3B base model, specifically tailored for the agency-ops domain. This model is designed to act as a sharp operator or advisor for AI-automation agencies, assisting with various operational tasks.

Key Capabilities

  • Lead Generation & Qualification: Assists with identifying and qualifying potential clients.
  • Discovery & Scoping: Helps in understanding client needs and defining project scope.
  • Pricing & Packaging: Supports the creation of pricing structures and service packages.
  • Proposals & SOWs: Aids in drafting professional proposals and Statements of Work.
  • Onboarding & Access: Facilitates client onboarding and access management.
  • Workflow Building & Shipping: Assists in developing and deploying AI agent workflows.
  • Tooling & Integrations: Provides guidance on relevant tools and system integrations.
  • Delivery & QA: Supports the delivery process and quality assurance.
  • Support & Retention: Helps with client support and retention strategies.
  • Agency Management: Offers insights into running the agency itself.

Provenance and Training

The model's v1.0 was trained on synthetic seed data generated by permissively-licensed local models (Apache/MIT teachers only), ensuring no distillation from closed models. Its niche capabilities are expected to sharpen further with consented real usage data flowing through the OpSpot flywheel. The model is available in safetensors (for transformers/vLLM) and model-q4_k_m.gguf (for llama.cpp/Ollama) formats.

Finetunes
1 models