beyoru/Ops-I

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 24, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Ops-I is a 27.8 billion parameter instruction-tuned model developed by beyoru, fine-tuned from Qwen3.6-27B. It is specifically designed as a workflow agent for business automation, excelling at tool discovery, chaining multiple tool calls, and discerning when not to act. The model demonstrates strong performance on the 600-task AutomationBench, particularly in finance and operations domains, with a validated context length of 16K tokens.

Loading preview...

Ops-I: A Specialized Workflow Agent for Business Automation

Ops-I is a 27.8 billion parameter model, fine-tuned from Qwen3.6-27B, specifically engineered to function as a workflow agent for business automation. Its core strength lies in processing business requests, identifying and chaining appropriate tools from a large set, and critically, understanding when to refrain from action. This model is particularly suited for complex, multi-step tool-use scenarios.

Key Capabilities & Performance

  • Business Workflow Automation: Optimized for agentic tasks involving tool use and decision-making in business contexts.
  • Strongest Open-Weight Performer: On the public 600-task AutomationBench, Ops-I achieved a 10.17% pass rate, outperforming its Qwen3.6-27B base and other open-weight models like gemma-4-31B-it.
  • Domain Expertise: Shows particular strength in finance and operations tasks, with 20 and 18 tasks passed respectively out of 100 in these domains.
  • Tool-Call Format: Utilizes Qwen3.5-style XML for tool calls (<function=...>), not JSON.
  • Context Length: Validated for effective use up to 16,384 tokens.

Known Limitations

The primary limitation identified is a tendency to over-act, meaning it is more prone to performing an action than to correctly refraining from one. For instance, it has a 53.7% success rate on positive assertions ("do send") but only 11.1% on negative assertions ("do not send"). Users are advised to implement confirmation gates for critical actions like sending emails or writing to sheets when deploying Ops-I in workflows.

When to Use Ops-I

  • Backend for Business Automation: Ideal for developers needing an open-weight model to power complex, multi-step automation workflows.
  • Tool-Use Scenarios: Excellent for applications requiring the model to discover, chain, and execute various tools based on user requests.
  • Specific Business Domains: Particularly effective for tasks within finance and operations.

Finetunes
1 models