Introducing Our Smallest, Most Efficient Model: Muse Ember

Anthropic introduced Muse Ember, its smallest and most efficient model to date, targeting repetitive agent workloads such as tool calls, coding subtasks, routing, extraction, and long document parsing. The company said Muse Ember matches Claude Haiku 5.5's best score at roughly one-third the cost and outperforms every setting of GPT-6 Luna, delivering more intelligence per dollar at any reasoning level. The launch matters because agentic workloads scale with volume, making per-task economics a key competitive battleground among model providers. Pricing is described as proposed and cost figures are estimates, so watch for finalized rates and independent benchmark confirmation.