OpenAI and Anthropic cut AI model costs with new flagship releases

Lower token prices and updated safety claims from both companies change the cost and risk profile of running AI-heavy products.

Anthropic and OpenAI have each launched new foundation models that target similar or better capability at significantly lower token prices, directly affecting what it costs teams to run AI-backed products. Anthropic’s Claude Opus 5.5 is positioned as a complex work and coding model, with claimed gains in agentic coding benchmarks over OpenAI’s GPT-6 Astra and a price cut to $4 per million input tokens and $20 per million output tokens from Opus 5’s $5 / $25. OpenAI’s new GPT-6 Sol and GPT-6 Luna models are trained with similar methods to GPT-6 Astra and are priced for efficiency, with Sol at $2 per million input and $10 per million output tokens, and Luna at $0.10 and $0.50 respectively. OpenAI reports that GPT-6 Sol makes about half as many factual mistakes as its predecessor and matches Fable 5.1’s coding performance at lower cost, while Anthropic says Opus 5.5 improves on difficult software, financial and business tasks and attempts to circumvent safety boundaries far less often. Developers can access Opus 5.5 via Claude and major cloud platforms, while GPT-6 Sol and Luna are available to paying ChatGPT Work and Codex customers, with Luna also reaching free and Go users in OpenAI’s desktop app.