In a recent LinkedIn post, Dan Sherrard-Smith offers a strategic framework for selecting the most appropriate Claude AI model to optimize token usage and cost-effectiveness. Sherrard-Smith, a proponent of efficient AI deployment, outlines a three-step decision tree designed to guide users through the complexities of choosing between Claude’s various iterations, from the swift Haiku to the powerful Fable.
The core of Sherrard-Smith’s advice centers on a systematic approach to task analysis. He emphasizes that understanding the nature and demands of a task is paramount to selecting the right tool. According to Dan Sherrard-Smith, the first crucial question is whether the task requires a complex answer. His decision tree suggests:
“No: use Haiku 4.5 or Sonnet 5. Yes: use Opus 4.8 or Fable 5 (if available)”
Understanding Task Complexity and Speed Requirements
Delving deeper into the decision-making process, Sherrard-Smith addresses the need for speed. He highlights that for tasks demanding immediate responses, the Haiku model is the optimal choice due to its lightweight nature and token-saving capabilities. This model is recommended for basic chat interactions without file attachments and can be enhanced with web search functionality. For slightly more involved, everyday tasks that still require quick turnaround, Sonnet 5 is presented as a robust option. Sherrard-Smith notes that Sonnet 5 is:
“Perfect for simple tasks. Connect your apps with Connectors: Slack, Google Drive, Notion, Figma and 50+ more.”
Conversely, if speed is not a primary concern, the focus shifts to the depth and ambition of the task. This leads to the second and third steps of Sherrard-Smith’s framework, which differentiate between deep work and the most challenging, ambitious projects.
Leveraging Opus and Fable for Intensive Workloads
For tasks that require significant effort but are not necessarily the user’s most critical or ambitious undertaking, Sherrard-Smith recommends the Opus 4.8 model. In his view, this model is best suited for deep work sessions, advising users to:
“Use Cowork. Put Effort on High. Always. Use Skills in Projects.”
He further suggests token-saving strategies for Opus, such as downloading files or converting work into Claude Skills after completion to prepare for a fresh session. This methodical approach aims to prevent unnecessary token consumption by re-reading entire conversation histories.
When the task at hand represents the pinnacle of ambition and complexity, Sherrard-Smith points to Fable 5 as the ultimate solution. He describes it as Anthropic’s smartest model, ideal for deep research and analytical decision-making. However, he issues a strong caution regarding its usage:
“It’s now pay-per-use (and the cost stacks up fast) … <10% of tasks actually need it … Use it 1-2 turns for strategy, then switch to Opus"
Sherrard-Smith’s overarching strategy is to use Fable 5 judiciously, primarily when Opus encounters limitations. The guiding principle, as articulated by Dan Sherrard-Smith, is to escalate to Fable only when necessary, routing all other tasks down the decision tree to conserve resources. He concludes by encouraging users to save and share this guide to prevent unnecessary token expenditure within their teams.
📝 About This Content
This article is based on insights shared by Dan Sherrard-Smith on LinkedIn.
📅 Originally posted on July 22, 2026 | View original post on LinkedIn →