Anthropic and OpenAI have released new AI models despite industry debate over slowing frontier development. Both companies say their latest systems offer comparable or improved performance at lower prices, targeting professional work, coding and enterprise use.

Anthropic’s Opus 5.5 is claimed to be stronger at complex tasks, software optimisation and financial analysis, while costing $4 per million input tokens and $20 per million output tokens. OpenAI’s GPT-6 Sol and Luna are advertised as up to 50% cheaper than earlier models, with Sol costing $2/$10 and Luna $0.10/$0.50 for input/output tokens; OpenAI also reports fewer factual and coding errors, faster responses and improved safety. Opus 5.5 is available through Claude and major cloud platforms, while the GPT-6 models are offered to selected ChatGPT Work and Codex customers.