Claude Opus 5.5 is now available. Anthropic responds to OpenAI with a cheaper and faster model
We have a real week of premieres in the AI field. On the same day as GPT-6 Sol and Luna, a new model from Anthropic – Claude Opus 5.5 – also debuts
On September 22, Anthropic showed the world a new AI model – Claude Opus 5.5. On the same day, OpenAI released the GPT-6 Sol and GPT-6 Luna models. According to Anthropic, the new Opus is comparable to Claude Fable 5.1, the company’s most powerful model from early September, in most tasks, and it costs less.
Cheaper and faster
API price list dropped by 20%. compared to Opus 5. A million input tokens now cost $4, and a million output tokens cost $20. Reading from the cache costs $0.20, i.e. 60 percent. less than before. Anthropic estimates that typical tasks will be up to 40 percent lower. cheaper. The model also accounts for approximately 30 percent. faster than its predecessor.
Strong in programming
In the Terminal-Bench 4.0 test, which is conducted by Stanford University and the Laude Institute to verify how AI programs and independently performs technical tasks, Opus 5.5 scored 66.4 percent.
OpenAI’s flagship model, GPT-6 Astra, achieved 57.9 percent, and Fable 5.1 achieved 55.8 percent.
In FrontierCode v1.1, a test that checks whether code written by AI is suitable for implementation, and not just whether it works, Opus 5.5 scored 54.4 percent, GPT-6 Astra 53.3 percent, and Fable 5.1 50.3 percent.
The new model also copes better than its predecessors with computer operation and office work. One of the testers transferred 680,000 with it. line of code in less than 24 hours. It would take the engineering team weeks.
Increased level of security
According to Anthropic Opus 5.5 by 85%. tries to circumvent imposed restrictions less often than Opus 5. It is also more resistant to prompt injection attacks and in this category it is the same as Fable 5.1. The company also introduced content marking (watermarking) due to the requirements of the EU AI Act.
Widely available from hyperscalers
The model runs on Claude apps on the Pro, Max, Team and Enterprise plans. It is also available via API and in AWS, Google Cloud and Microsoft Azure clouds. Subscribers got higher usage limits. Claude Sonnet 5.5 and Claude Haiku 5.5 are scheduled to debut in the coming weeks.
Sources: Antyweb, Anthropic, MacRumors, TechRepublic, TechCrunch
