Anthropic, OpenAI race to cheaper mid-tier models

Anthropic and OpenAI released their cheaper mid-tier models within weeks of each other, with safety routing and pricing as the main selling points. Anthropic's Claude Opus 5.5 enters as the first release since CEO Dario Amodei said the company would pace the frontier, and it inherits safeguards from its Fable line. OpenAI's Sol and Luna, released 90 minutes after Opus 5.5, target different work: Sol for complex tasks like coding, Luna for high-volume clerical work. Both are priced at half the 5.6 series, which OpenAI credits to caching and inference improvements.

The pricing moves come as Anthropic also released Sonnet 5.5, its mid-tier model that it says is 30% faster than Sonnet 5 with a slower token burn rate. Anthropic's benchmarks show Sonnet 5.5 beating Opus 5.5 on agentic coding, which the company attributes to spawning multiple agents within cost limits. A new Haiku is planned in the coming weeks. A week after the Sol and Luna release, OpenAI showed GPT-6.1 Sol at DevDay, which it says nears GPT-6 Astra on agentic coding at one-fifth the token prices, with factual errors falling from 11.4% to 7.7% at low reasoning effort.

The outcome hinges on whether these cost and error improvements hold up outside internal benchmarks, and on how buyers weigh price against the safeguards both companies are now emphasizing. For teams running high-volume or agentic workloads, the gap between mid-tier and frontier pricing may matter more than maximum capability, though the article does not specify how these models compare on third-party evaluations.