Anthropic and OpenAI have announced new versions of their flagship models, Opus 5.5 and GPT-6 Astra, with performance similar or superior to predecessors but at lower per-token prices. The information was published by Engadget and reflects the ongoing price war in generative AI.
Benchmarks released by Anthropic show Opus 5.5 beat GPT-6 Astra on Terminal-Bench 4.0 and FrontierCode v1.1 (Main), indicating an edge in agentic programming tasks. The company also highlights lower per-token costs: $4 for input tokens and $20 for output tokens, compared to $5 and an undisclosed figure from the competitor.
The cost reduction did not come at the expense of general quality. Both models maintain the reasoning and code-generation capabilities of earlier versions but with greater computational efficiency. That means developers can run more API calls for the same budget.
For teams operating AI agent systems, the news has direct implications for operational costs and provider choice. Anthropic's advantage in agentic coding may shift architectural decisions, especially in development automation and code review pipelines.
The Week in Tech sees the move as confirming the commoditization trend among frontier models. Price differences and performance on specific tasks, such as programming, should weigh more heavily on vendor selection than generic accuracy metrics.