What is the Claude advisor strategy?

A pattern where Opus serves as an on-demand advisor to a cheaper executor model (Sonnet or Haiku), delivering near-Opus intelligence at a fraction of the cost via a single API tool declaration.

How much does the advisor strategy save?

On SWE-bench Multilingual, it reduced cost by 11.9% while improving performance. The Haiku + Opus combo costs 85% less than Sonnet alone.

How do you implement the Claude advisor strategy?

Add advisor_20260301 as a tool declaration in the Messages API request with the beta header anthropic-beta: advisor-tool-2026-03-01.

What performance gains does the advisor strategy provide?

Sonnet + Opus advisor gained 2.7 percentage points on SWE-bench Multilingual. Haiku + Opus advisor more than doubled its BrowseComp score from 19.7% to 41.2%.

Can Haiku use Opus as an advisor?

Yes. Haiku with Opus advisor scored 41.2% on BrowseComp vs 19.7% solo, while costing 85% less than running Sonnet.

← Back to dashboard

clawsmith.com/signal/anthropic-advisor-strategy-claude-platform

🔥 HypeWide OpenLive

Anthropic Launches Advisor Strategy on Claude Platform

Anthropic introduced the advisor strategy to the Claude Platform, letting developers pair Opus as an on-demand advisor with Sonnet or Haiku as the executor. On SWE-bench Multilingual, Sonnet + Opus advisor beat Sonnet alone by 2.7 percentage points while costing 11.9% less. On BrowseComp, Haiku + Opus advisor scored 41.2% vs 19.7% solo, more than doubling performance at 85% less cost than Sonnet. One-line API change via advisor_20260301.

Product Idea from this Signal

A routing middleware that pairs an expensive advisor model with a cheap executor model for OpenClaw agents, cutting API costs by 80% while maintaining output quality

2230.5k ▲

OpenClaw agents burning $50-200/month on premium models for every task, even simple ones. Anthropic proved the advisor/executor pattern works with 2.2M engagement on X when they announced Claude's advisor strategy pairing Opus with Sonnet/Haiku. Existing routing tools pick the cheapest model per task, but none implement a two-tier system where a frontier model reviews and steers a budget model's reasoning. This middleware intercepts OpenClaw agent requests, routes the planning step to an expensive advisor (Opus, GPT-5.4), then hands execution to a budget model (Haiku, Flash) with the advisor's structured guidance attached.

CLIOPEN-SOURCECOST-OPTIMIZATIONMODEL-ROUTINGDEVTOOL

CompetitiveView Opportunity →