Anthropic Launches Claude Opus 5.5: Cheaper, Faster, and Safer Than Predecessors

Key points

• Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens, a 20% price cut from Opus 5.
• The model achieves a score of 58 on the Artificial Analysis Intelligence Index, the highest recorded, surpassing GPT-6 Astra and Fable 5.1.
• Anthropic introduced 'Preserved Thinking' to prevent distillation attacks and added watermarking for EU AI Act compliance.
• OpenAI launched GPT-6 Sol and Luna, offering improved alignment and lower rates of unauthorized actions compared to previous versions.
• Microsoft Azure made GPT-6 Astra, Sol, and Luna available in Foundry, emphasizing cost-per-task efficiency for enterprise agents.

Background

The release of Claude Opus 5.5 follows a period of intense competition between AI providers, where cost efficiency and safety have become primary differentiators. Previous models, such as Opus 5 and Fable 5.1, faced criticism for high operating costs and formulaic writing styles, often referred to as 'Claudish.' The industry has also seen a push for stricter safety standards, with OpenAI and Anthropic both emphasizing alignment and third-party evaluations to mitigate risks associated with increasingly capable models.

How outlets are covering it

Anthropic emphasizes the cost-effectiveness and improved communication quality of Opus 5.5, highlighting its ability to match Fable 5.1 performance at a lower price. The Hacker News focuses on the safety improvements, noting that Opus 5.5 exhibits less misaligned behavior and is more resistant to prompt injection. Microsoft Azure highlights the practical integration of OpenAI's GPT-6 models into its Foundry platform, stressing the importance of cost-per-task metrics for enterprise customers. While Anthropic and OpenAI both prioritize safety, their approaches differ, with Anthropic introducing specific anti-distillation measures and OpenAI focusing on third-party evaluations.

Why it matters

The launch of Claude Opus 5.5 and the expansion of OpenAI's GPT-6 lineup signal a shift in the AI industry towards more efficient, safer, and cost-effective models. This development is crucial for enterprises looking to integrate AI into their workflows without incurring prohibitive costs. Additionally, the emphasis on safety and compliance reflects growing regulatory pressures and the need for responsible AI development.

What to watch

Anthropic plans to release Claude Sonnet 5.5 and Haiku 5.5 in the coming weeks, with similar gains in performance and efficiency. OpenAI will continue to support third-party evaluations of its models, and Microsoft Azure will likely expand its integration of GPT-6 models into more enterprise solutions. The industry will also see further developments in AI safety standards and regulatory compliance as models become more capable.