Anthropic Optimizes Fable 5 with Sonnet 5 to Cut Costs
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Anthropic Optimizes Fable 5 with Sonnet 5 to Reduce Costs
Claude Fable 5 is expensive. Anthropic now recommends using it primarily as a planner, delegating execution to smaller models. The Claude development team presents two strategies. In the "Advisor" model, Sonnet 5 acts as the executor and only calls on Fable 5 when it needs advice. On SWE-bench Pro, this combination achieves about 92% of Fable 5's solo performance for 63% of the cost, according to Anthropic. Fable 5 is called upon about once per task.
- Advisor Model: Sonnet 5 does the work and consults Fable 5 only when necessary.
In the second model, Fable 5 acts as a planner that delegates tasks to Sonnet 5 work agents. On BrowseComp, this provides 96% of Fable 5's performance for 46% of the cost. Both models operate via Claude Managed Agents, with each sub-agent using its own cache to avoid duplicate context costs.
- Orchestrator Model: Fable 5 plans and distributes tasks to multiple Sonnet 5 workers.
Anthropic is likely sharing this advice due to increasing price pressure. Chinese open-source models are already starting to compete with Western prices, and the new GPT-5.6 Sol is significantly cheaper per token and is also expected to be more efficient in terms of tokens.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.