Anthropic Launches Advisor Strategy for the Claude API
On April 9, 2026, Anthropic released the Advisor Strategy on the Claude Platform. The feature lets developers pair Claude Opus 4.6 as a strategic advisor with Sonnet 4.6 or Haiku 4.5 as the primary executor in agentic workflows.
How It Works
The executor model (Sonnet or Haiku) runs the task end to end. It calls tools, reads results, and works toward a solution. When it hits a decision it cannot resolve on its own, it consults Opus. Opus reviews the shared context and returns a plan, a correction, or a stop signal. The executor then continues.
Developers access it by adding an advisor tool entry with type "advisor_20260301" and model "claude-opus-4-6" to their tools list in the Messages API. The beta header "anthropic-beta: advisor-tool-2026-03-01" is required.
Performance Results
Sonnet with Opus as advisor scored 2.7 percentage points higher on SWE-bench Multilingual compared to Sonnet alone, while reducing cost per agentic task by 11.9%. Haiku with Opus as advisor scored 41.2% on BrowseComp, more than double its solo score of 19.7%.
Why It Matters
The Advisor Strategy addresses a common tradeoff in AI agent design: using the smartest model for every step is expensive, but using a cheaper model everywhere lowers quality. This approach lets developers get near-Opus results at a fraction of the cost by only consulting the larger model when needed.
Discussion
Sign in to comment. Your account must be at least 1 day old.