Anthropic Unveils Claude Opus 5.5, Promising Lower Costs and Competitive Coding Performance
Anthropic has released Claude Opus 5.5, a new flagship AI model priced at $4 per million input tokens and $20 per million output tokens. The company claims it outperforms OpenAI's GPT-5.6 Sol on a software development benchmark while costing about a third as much to run. The model also features a 1-million-token context window and improved safety controls.
Claude Opus 5.5 enters the market at $4 per million input tokens and $20 per million output tokens, a 20% reduction from its predecessor, with cache reads dropping 60% to $0.20 per million tokens. Anthropic reports output generation runs over 30% faster than Opus 5. The model carries a 1-million-token context window and underwent pre-release evaluation by external safety groups Frontier Design and METR, achieving the company's strongest automated behavioral audit results to date.
The model is available through Anthropic directly and via Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic also raised five-hour usage limits across Pro, Max, Team, and Enterprise subscription tiers. The launch follows CEO Dario Amodei's recent public call for the global AI community to slow the release of new capabilities, creating an unusual dynamic where the company simultaneously advances its frontier model while urging caution industry-wide.
This launch could reshape enterprise AI purchasing decisions, as organizations weighing model choices may prioritize cost efficiency alongside raw capability. The pricing pressure on competitors may accelerate a broader trend toward performance-per-dollar as a primary metric. Businesses deploying AI at scale could see meaningful operational savings, while the safety-testing emphasis may influence public trust in frontier models. However, benchmark claims warrant scrutiny, as real-world results often diverge from controlled evaluations, and smaller competitors could face heightened difficulty matching Anthropic's combination of price and performance.